Merge "remove sad x3,x8 specializations" into main

20 files changed, 33 insertions, 2123 deletions

diff --git a/test/sad_test.cc b/test/sad_test.cc
index ee10a4638..560c5f382 100644
--- a/test/sad_test.cc
+++ b/test/sad_test.cc
@@ -56,8 +56,6 @@ typedef void (*SadMxNx8Func)(const uint8_t *src_ptr, int src_stride,
                              const uint8_t *ref_ptr, int ref_stride,
                              unsigned int *sad_array);
 
-typedef TestParams<SadMxNx8Func> SadMxNx8Param;
-
 using libvpx_test::ACMRandom;
 
 namespace {
@@ -266,30 +264,6 @@ class SADTestBase : public ::testing::TestWithParam<ParamType> {
   ParamType params_;
 };
 
-class SADx8Test : public SADTestBase<SadMxNx8Param> {
- public:
-  SADx8Test() : SADTestBase(GetParam()) {}
-
- protected:
-  void SADs(unsigned int *results) const {
-    const uint8_t *reference = GetReferenceFromOffset(0);
-
-    ASM_REGISTER_STATE_CHECK(params_.func(
-        source_data_, source_stride_, reference, reference_stride_, results));
-  }
-
-  void CheckSADs() const {
-    uint32_t reference_sad;
-    DECLARE_ALIGNED(kDataAlignment, uint32_t, exp_sad[8]);
-
-    SADs(exp_sad);
-    for (int offset = 0; offset < 8; ++offset) {
-      reference_sad = ReferenceSAD(offset);
-      EXPECT_EQ(reference_sad, exp_sad[offset]) << "offset " << offset;
-    }
-  }
-};
-
 class SADx4Test : public SADTestBase<SadMxNx4Param> {
  public:
   SADx4Test() : SADTestBase(GetParam()) {}
@@ -564,13 +538,6 @@ TEST_P(SADx4Test, DISABLED_Speed) {
   reference_stride_ = tmp_stride;
 }
 
-TEST_P(SADx8Test, Regular) {
-  FillRandomWH(source_data_, source_stride_, params_.width, params_.height);
-  FillRandomWH(GetReferenceFromOffset(0), reference_stride_, params_.width + 8,
-               params_.height);
-  CheckSADs();
-}
-
 //------------------------------------------------------------------------------
 // C functions
 const SadMxNParam c_tests[] = {
@@ -747,24 +714,6 @@ const SadMxNx4Param x4d_c_tests[] = {
 };
 INSTANTIATE_TEST_SUITE_P(C, SADx4Test, ::testing::ValuesIn(x4d_c_tests));
 
-// TODO(angiebird): implement the marked-down sad functions
-const SadMxNx8Param x8_c_tests[] = {
-  // SadMxNx8Param(64, 64, &vpx_sad64x64x8_c),
-  // SadMxNx8Param(64, 32, &vpx_sad64x32x8_c),
-  // SadMxNx8Param(32, 64, &vpx_sad32x64x8_c),
-  SadMxNx8Param(32, 32, &vpx_sad32x32x8_c),
-  // SadMxNx8Param(32, 16, &vpx_sad32x16x8_c),
-  // SadMxNx8Param(16, 32, &vpx_sad16x32x8_c),
-  SadMxNx8Param(16, 16, &vpx_sad16x16x8_c),
-  SadMxNx8Param(16, 8, &vpx_sad16x8x8_c),
-  SadMxNx8Param(8, 16, &vpx_sad8x16x8_c),
-  SadMxNx8Param(8, 8, &vpx_sad8x8x8_c),
-  // SadMxNx8Param(8, 4, &vpx_sad8x4x8_c),
-  // SadMxNx8Param(4, 8, &vpx_sad4x8x8_c),
-  SadMxNx8Param(4, 4, &vpx_sad4x4x8_c),
-};
-INSTANTIATE_TEST_SUITE_P(C, SADx8Test, ::testing::ValuesIn(x8_c_tests));
-
 //------------------------------------------------------------------------------
 // ARM functions
 #if HAVE_NEON
@@ -992,18 +941,6 @@ INSTANTIATE_TEST_SUITE_P(SSE2, SADx4Test, ::testing::ValuesIn(x4d_sse2_tests));
 // Only functions are x3, which do not have tests.
 #endif  // HAVE_SSSE3
 
-#if HAVE_SSE4_1
-const SadMxNx8Param x8_sse4_1_tests[] = {
-  SadMxNx8Param(16, 16, &vpx_sad16x16x8_sse4_1),
-  SadMxNx8Param(16, 8, &vpx_sad16x8x8_sse4_1),
-  SadMxNx8Param(8, 16, &vpx_sad8x16x8_sse4_1),
-  SadMxNx8Param(8, 8, &vpx_sad8x8x8_sse4_1),
-  SadMxNx8Param(4, 4, &vpx_sad4x4x8_sse4_1),
-};
-INSTANTIATE_TEST_SUITE_P(SSE4_1, SADx8Test,
-                         ::testing::ValuesIn(x8_sse4_1_tests));
-#endif  // HAVE_SSE4_1
-
 #if HAVE_AVX2
 const SadMxNParam avx2_tests[] = {
   SadMxNParam(64, 64, &vpx_sad64x64_avx2),
@@ -1029,11 +966,6 @@ const SadMxNx4Param x4d_avx2_tests[] = {
 };
 INSTANTIATE_TEST_SUITE_P(AVX2, SADx4Test, ::testing::ValuesIn(x4d_avx2_tests));
 
-const SadMxNx8Param x8_avx2_tests[] = {
-  // SadMxNx8Param(64, 64, &vpx_sad64x64x8_c),
-  SadMxNx8Param(32, 32, &vpx_sad32x32x8_avx2),
-};
-INSTANTIATE_TEST_SUITE_P(AVX2, SADx8Test, ::testing::ValuesIn(x8_avx2_tests));
 #endif  // HAVE_AVX2
 
 #if HAVE_AVX512
diff --git a/vp8/common/rtcd_defs.pl b/vp8/common/rtcd_defs.pl
index 32601b4eb..c7911032f 100644
--- a/vp8/common/rtcd_defs.pl
+++ b/vp8/common/rtcd_defs.pl
@@ -216,11 +216,6 @@ specialize qw/vp8_mbuverror sse2 msa/;
 #
 # Motion search
 #
-add_proto qw/int vp8_full_search_sad/, "struct macroblock *x, struct block *b, struct blockd *d, union int_mv *ref_mv, int sad_per_bit, int distance, struct variance_vtable *fn_ptr, int *mvcost[2], union int_mv *center_mv";
-specialize qw/vp8_full_search_sad sse3 sse4_1/;
-$vp8_full_search_sad_sse3=vp8_full_search_sadx3;
-$vp8_full_search_sad_sse4_1=vp8_full_search_sadx8;
-
 add_proto qw/int vp8_refining_search_sad/, "struct macroblock *x, struct block *b, struct blockd *d, union int_mv *ref_mv, int error_per_bit, int search_range, struct variance_vtable *fn_ptr, int *mvcost[2], union int_mv *center_mv";
 specialize qw/vp8_refining_search_sad sse2 msa/;
 $vp8_refining_search_sad_sse2=vp8_refining_search_sadx4;
diff --git a/vp8/encoder/mcomp.c b/vp8/encoder/mcomp.c
index 4ab6c7b3d..769c2f558 100644
--- a/vp8/encoder/mcomp.c
+++ b/vp8/encoder/mcomp.c
@@ -1280,10 +1280,10 @@ int vp8_diamond_search_sadx4(MACROBLOCK *x, BLOCK *b, BLOCKD *d, int_mv *ref_mv,
 }
 #endif  // HAVE_SSE2 || HAVE_MSA
 
-int vp8_full_search_sad_c(MACROBLOCK *x, BLOCK *b, BLOCKD *d, int_mv *ref_mv,
-                          int sad_per_bit, int distance,
-                          vp8_variance_fn_ptr_t *fn_ptr, int *mvcost[2],
-                          int_mv *center_mv) {
+int vp8_full_search_sad(MACROBLOCK *x, BLOCK *b, BLOCKD *d, int_mv *ref_mv,
+                        int sad_per_bit, int distance,
+                        vp8_variance_fn_ptr_t *fn_ptr, int *mvcost[2],
+                        int_mv *center_mv) {
   unsigned char *what = (*(b->base_src) + b->src);
   int what_stride = b->src_stride;
   unsigned char *in_what;
@@ -1327,217 +1327,6 @@ int vp8_full_search_sad_c(MACROBLOCK *x, BLOCK *b, BLOCKD *d, int_mv *ref_mv,
   bestsad = fn_ptr->sdf(what, what_stride, bestaddress, in_what_stride) +
             mvsad_err_cost(best_mv, &fcenter_mv, mvsadcost, sad_per_bit);
 
-  /* Apply further limits to prevent us looking using vectors that
-   * stretch beyiond the UMV border
-   */
-  if (col_min < x->mv_col_min) col_min = x->mv_col_min;
-
-  if (col_max > x->mv_col_max) col_max = x->mv_col_max;
-
-  if (row_min < x->mv_row_min) row_min = x->mv_row_min;
-
-  if (row_max > x->mv_row_max) row_max = x->mv_row_max;
-
-  for (r = row_min; r < row_max; ++r) {
-    this_mv.as_mv.row = r;
-    check_here = r * mv_stride + in_what + col_min;
-
-    for (c = col_min; c < col_max; ++c) {
-      thissad = fn_ptr->sdf(what, what_stride, check_here, in_what_stride);
-
-      this_mv.as_mv.col = c;
-      thissad += mvsad_err_cost(&this_mv, &fcenter_mv, mvsadcost, sad_per_bit);
-
-      if (thissad < bestsad) {
-        bestsad = thissad;
-        best_mv->as_mv.row = r;
-        best_mv->as_mv.col = c;
-        bestaddress = check_here;
-      }
-
-      check_here++;
-    }
-  }
-
-  this_mv.as_mv.row = best_mv->as_mv.row << 3;
-  this_mv.as_mv.col = best_mv->as_mv.col << 3;
-
-  return fn_ptr->vf(what, what_stride, bestaddress, in_what_stride, &thissad) +
-         mv_err_cost(&this_mv, center_mv, mvcost, x->errorperbit);
-}
-
-#if HAVE_SSSE3
-int vp8_full_search_sadx3(MACROBLOCK *x, BLOCK *b, BLOCKD *d, int_mv *ref_mv,
-                          int sad_per_bit, int distance,
-                          vp8_variance_fn_ptr_t *fn_ptr, int *mvcost[2],
-                          int_mv *center_mv) {
-  unsigned char *what = (*(b->base_src) + b->src);
-  int what_stride = b->src_stride;
-  unsigned char *in_what;
-  int pre_stride = x->e_mbd.pre.y_stride;
-  unsigned char *base_pre = x->e_mbd.pre.y_buffer;
-  int in_what_stride = pre_stride;
-  int mv_stride = pre_stride;
-  unsigned char *bestaddress;
-  int_mv *best_mv = &d->bmi.mv;
-  int_mv this_mv;
-  unsigned int bestsad;
-  unsigned int thissad;
-  int r, c;
-
-  unsigned char *check_here;
-
-  int ref_row = ref_mv->as_mv.row;
-  int ref_col = ref_mv->as_mv.col;
-
-  int row_min = ref_row - distance;
-  int row_max = ref_row + distance;
-  int col_min = ref_col - distance;
-  int col_max = ref_col + distance;
-
-  unsigned int sad_array[3];
-
-  int *mvsadcost[2];
-  int_mv fcenter_mv;
-
-  mvsadcost[0] = x->mvsadcost[0];
-  mvsadcost[1] = x->mvsadcost[1];
-  fcenter_mv.as_mv.row = center_mv->as_mv.row >> 3;
-  fcenter_mv.as_mv.col = center_mv->as_mv.col >> 3;
-
-  /* Work out the mid point for the search */
-  in_what = base_pre + d->offset;
-  bestaddress = in_what + (ref_row * pre_stride) + ref_col;
-
-  best_mv->as_mv.row = ref_row;
-  best_mv->as_mv.col = ref_col;
-
-  /* Baseline value at the centre */
-  bestsad = fn_ptr->sdf(what, what_stride, bestaddress, in_what_stride) +
-            mvsad_err_cost(best_mv, &fcenter_mv, mvsadcost, sad_per_bit);
-
-  /* Apply further limits to prevent us looking using vectors that stretch
-   * beyond the UMV border
-   */
-  if (col_min < x->mv_col_min) col_min = x->mv_col_min;
-
-  if (col_max > x->mv_col_max) col_max = x->mv_col_max;
-
-  if (row_min < x->mv_row_min) row_min = x->mv_row_min;
-
-  if (row_max > x->mv_row_max) row_max = x->mv_row_max;
-
-  for (r = row_min; r < row_max; ++r) {
-    this_mv.as_mv.row = r;
-    check_here = r * mv_stride + in_what + col_min;
-    c = col_min;
-
-    while ((c + 2) < col_max) {
-      int i;
-
-      fn_ptr->sdx3f(what, what_stride, check_here, in_what_stride, sad_array);
-
-      for (i = 0; i < 3; ++i) {
-        thissad = sad_array[i];
-
-        if (thissad < bestsad) {
-          this_mv.as_mv.col = c;
-          thissad +=
-              mvsad_err_cost(&this_mv, &fcenter_mv, mvsadcost, sad_per_bit);
-
-          if (thissad < bestsad) {
-            bestsad = thissad;
-            best_mv->as_mv.row = r;
-            best_mv->as_mv.col = c;
-            bestaddress = check_here;
-          }
-        }
-
-        check_here++;
-        c++;
-      }
-    }
-
-    while (c < col_max) {
-      thissad = fn_ptr->sdf(what, what_stride, check_here, in_what_stride);
-
-      if (thissad < bestsad) {
-        this_mv.as_mv.col = c;
-        thissad +=
-            mvsad_err_cost(&this_mv, &fcenter_mv, mvsadcost, sad_per_bit);
-
-        if (thissad < bestsad) {
-          bestsad = thissad;
-          best_mv->as_mv.row = r;
-          best_mv->as_mv.col = c;
-          bestaddress = check_here;
-        }
-      }
-
-      check_here++;
-      c++;
-    }
-  }
-
-  this_mv.as_mv.row = best_mv->as_mv.row << 3;
-  this_mv.as_mv.col = best_mv->as_mv.col << 3;
-
-  return fn_ptr->vf(what, what_stride, bestaddress, in_what_stride, &thissad) +
-         mv_err_cost(&this_mv, center_mv, mvcost, x->errorperbit);
-}
-#endif  // HAVE_SSSE3
-
-#if HAVE_SSE4_1
-int vp8_full_search_sadx8(MACROBLOCK *x, BLOCK *b, BLOCKD *d, int_mv *ref_mv,
-                          int sad_per_bit, int distance,
-                          vp8_variance_fn_ptr_t *fn_ptr, int *mvcost[2],
-                          int_mv *center_mv) {
-  unsigned char *what = (*(b->base_src) + b->src);
-  int what_stride = b->src_stride;
-  int pre_stride = x->e_mbd.pre.y_stride;
-  unsigned char *base_pre = x->e_mbd.pre.y_buffer;
-  unsigned char *in_what;
-  int in_what_stride = pre_stride;
-  int mv_stride = pre_stride;
-  unsigned char *bestaddress;
-  int_mv *best_mv = &d->bmi.mv;
-  int_mv this_mv;
-  unsigned int bestsad;
-  unsigned int thissad;
-  int r, c;
-
-  unsigned char *check_here;
-
-  int ref_row = ref_mv->as_mv.row;
-  int ref_col = ref_mv->as_mv.col;
-
-  int row_min = ref_row - distance;
-  int row_max = ref_row + distance;
-  int col_min = ref_col - distance;
-  int col_max = ref_col + distance;
-
-  DECLARE_ALIGNED(16, unsigned int, sad_array8[8]);
-  unsigned int sad_array[3];
-
-  int *mvsadcost[2];
-  int_mv fcenter_mv;
-
-  mvsadcost[0] = x->mvsadcost[0];
-  mvsadcost[1] = x->mvsadcost[1];
-  fcenter_mv.as_mv.row = center_mv->as_mv.row >> 3;
-  fcenter_mv.as_mv.col = center_mv->as_mv.col >> 3;
-
-  /* Work out the mid point for the search */
-  in_what = base_pre + d->offset;
-  bestaddress = in_what + (ref_row * pre_stride) + ref_col;
-
-  best_mv->as_mv.row = ref_row;
-  best_mv->as_mv.col = ref_col;
-
-  /* Baseline value at the centre */
-  bestsad = fn_ptr->sdf(what, what_stride, bestaddress, in_what_stride) +
-            mvsad_err_cost(best_mv, &fcenter_mv, mvsadcost, sad_per_bit);
-
   /* Apply further limits to prevent us looking using vectors that stretch
    * beyond the UMV border
    */
@@ -1552,61 +1341,8 @@ int vp8_full_search_sadx8(MACROBLOCK *x, BLOCK *b, BLOCKD *d, int_mv *ref_mv,
   for (r = row_min; r < row_max; ++r) {
     this_mv.as_mv.row = r;
     check_here = r * mv_stride + in_what + col_min;
-    c = col_min;
-
-    while ((c + 7) < col_max) {
-      int i;
-
-      fn_ptr->sdx8f(what, what_stride, check_here, in_what_stride, sad_array8);
-
-      for (i = 0; i < 8; ++i) {
-        thissad = sad_array8[i];
-
-        if (thissad < bestsad) {
-          this_mv.as_mv.col = c;
-          thissad +=
-              mvsad_err_cost(&this_mv, &fcenter_mv, mvsadcost, sad_per_bit);
-
-          if (thissad < bestsad) {
-            bestsad = thissad;
-            best_mv->as_mv.row = r;
-            best_mv->as_mv.col = c;
-            bestaddress = check_here;
-          }
-        }
-
-        check_here++;
-        c++;
-      }
-    }
-
-    while ((c + 2) < col_max) {
-      int i;
-
-      fn_ptr->sdx3f(what, what_stride, check_here, in_what_stride, sad_array);
-
-      for (i = 0; i < 3; ++i) {
-        thissad = sad_array[i];
 
-        if (thissad < bestsad) {
-          this_mv.as_mv.col = c;
-          thissad +=
-              mvsad_err_cost(&this_mv, &fcenter_mv, mvsadcost, sad_per_bit);
-
-          if (thissad < bestsad) {
-            bestsad = thissad;
-            best_mv->as_mv.row = r;
-            best_mv->as_mv.col = c;
-            bestaddress = check_here;
-          }
-        }
-
-        check_here++;
-        c++;
-      }
-    }
-
-    while (c < col_max) {
+    for (c = col_min; c < col_max; ++c) {
       thissad = fn_ptr->sdf(what, what_stride, check_here, in_what_stride);
 
       if (thissad < bestsad) {
@@ -1623,7 +1359,6 @@ int vp8_full_search_sadx8(MACROBLOCK *x, BLOCK *b, BLOCKD *d, int_mv *ref_mv,
       }
 
       check_here++;
-      c++;
     }
   }
 
@@ -1633,7 +1368,6 @@ int vp8_full_search_sadx8(MACROBLOCK *x, BLOCK *b, BLOCKD *d, int_mv *ref_mv,
   return fn_ptr->vf(what, what_stride, bestaddress, in_what_stride, &thissad) +
          mv_err_cost(&this_mv, center_mv, mvcost, x->errorperbit);
 }
-#endif  // HAVE_SSE4_1
 
 int vp8_refining_search_sad_c(MACROBLOCK *x, BLOCK *b, BLOCKD *d,
                               int_mv *ref_mv, int error_per_bit,
diff --git a/vp8/encoder/mcomp.h b/vp8/encoder/mcomp.h
index 57c18f523..1ee6fe5dd 100644
--- a/vp8/encoder/mcomp.h
+++ b/vp8/encoder/mcomp.h
@@ -50,10 +50,10 @@ fractional_mv_step_fp vp8_find_best_sub_pixel_step;
 fractional_mv_step_fp vp8_find_best_half_pixel_step;
 fractional_mv_step_fp vp8_skip_fractional_mv_step;
 
-typedef int (*vp8_full_search_fn_t)(MACROBLOCK *x, BLOCK *b, BLOCKD *d,
-                                    int_mv *ref_mv, int sad_per_bit,
-                                    int distance, vp8_variance_fn_ptr_t *fn_ptr,
-                                    int *mvcost[2], int_mv *center_mv);
+int vp8_full_search_sad(MACROBLOCK *x, BLOCK *b, BLOCKD *d, int_mv *ref_mv,
+                        int sad_per_bit, int distance,
+                        vp8_variance_fn_ptr_t *fn_ptr, int *mvcost[2],
+                        int_mv *center_mv);
 
 typedef int (*vp8_refining_search_fn_t)(MACROBLOCK *x, BLOCK *b, BLOCKD *d,
                                         int_mv *ref_mv, int sad_per_bit,
diff --git a/vp8/encoder/onyx_if.c b/vp8/encoder/onyx_if.c
index f09177c7f..ffb3867dd 100644
--- a/vp8/encoder/onyx_if.c
+++ b/vp8/encoder/onyx_if.c
@@ -2012,36 +2012,26 @@ struct VP8_COMP *vp8_create_compressor(VP8_CONFIG *oxcf) {
   cpi->fn_ptr[BLOCK_16X16].sdf = vpx_sad16x16;
   cpi->fn_ptr[BLOCK_16X16].vf = vpx_variance16x16;
   cpi->fn_ptr[BLOCK_16X16].svf = vpx_sub_pixel_variance16x16;
-  cpi->fn_ptr[BLOCK_16X16].sdx3f = vpx_sad16x16x3;
-  cpi->fn_ptr[BLOCK_16X16].sdx8f = vpx_sad16x16x8;
   cpi->fn_ptr[BLOCK_16X16].sdx4df = vpx_sad16x16x4d;
 
   cpi->fn_ptr[BLOCK_16X8].sdf = vpx_sad16x8;
   cpi->fn_ptr[BLOCK_16X8].vf = vpx_variance16x8;
   cpi->fn_ptr[BLOCK_16X8].svf = vpx_sub_pixel_variance16x8;
-  cpi->fn_ptr[BLOCK_16X8].sdx3f = vpx_sad16x8x3;
-  cpi->fn_ptr[BLOCK_16X8].sdx8f = vpx_sad16x8x8;
   cpi->fn_ptr[BLOCK_16X8].sdx4df = vpx_sad16x8x4d;
 
   cpi->fn_ptr[BLOCK_8X16].sdf = vpx_sad8x16;
   cpi->fn_ptr[BLOCK_8X16].vf = vpx_variance8x16;
   cpi->fn_ptr[BLOCK_8X16].svf = vpx_sub_pixel_variance8x16;
-  cpi->fn_ptr[BLOCK_8X16].sdx3f = vpx_sad8x16x3;
-  cpi->fn_ptr[BLOCK_8X16].sdx8f = vpx_sad8x16x8;
   cpi->fn_ptr[BLOCK_8X16].sdx4df = vpx_sad8x16x4d;
 
   cpi->fn_ptr[BLOCK_8X8].sdf = vpx_sad8x8;
   cpi->fn_ptr[BLOCK_8X8].vf = vpx_variance8x8;
   cpi->fn_ptr[BLOCK_8X8].svf = vpx_sub_pixel_variance8x8;
-  cpi->fn_ptr[BLOCK_8X8].sdx3f = vpx_sad8x8x3;
-  cpi->fn_ptr[BLOCK_8X8].sdx8f = vpx_sad8x8x8;
   cpi->fn_ptr[BLOCK_8X8].sdx4df = vpx_sad8x8x4d;
 
   cpi->fn_ptr[BLOCK_4X4].sdf = vpx_sad4x4;
   cpi->fn_ptr[BLOCK_4X4].vf = vpx_variance4x4;
   cpi->fn_ptr[BLOCK_4X4].svf = vpx_sub_pixel_variance4x4;
-  cpi->fn_ptr[BLOCK_4X4].sdx3f = vpx_sad4x4x3;
-  cpi->fn_ptr[BLOCK_4X4].sdx8f = vpx_sad4x4x8;
   cpi->fn_ptr[BLOCK_4X4].sdx4df = vpx_sad4x4x4d;
 
 #if VPX_ARCH_X86 || VPX_ARCH_X86_64
@@ -2052,7 +2042,6 @@ struct VP8_COMP *vp8_create_compressor(VP8_CONFIG *oxcf) {
   cpi->fn_ptr[BLOCK_4X4].copymem = vp8_copy32xn;
 #endif
 
-  cpi->full_search_sad = vp8_full_search_sad;
   cpi->diamond_search_sad = vp8_diamond_search_sad;
   cpi->refining_search_sad = vp8_refining_search_sad;
 
diff --git a/vp8/encoder/onyx_int.h b/vp8/encoder/onyx_int.h
index 7f8298e44..424f51b18 100644
--- a/vp8/encoder/onyx_int.h
+++ b/vp8/encoder/onyx_int.h
@@ -549,7 +549,6 @@ typedef struct VP8_COMP {
   unsigned char *partition_d_end[MAX_PARTITIONS];
 
   fractional_mv_step_fp *find_fractional_mv_step;
-  vp8_full_search_fn_t full_search_sad;
   vp8_refining_search_fn_t refining_search_sad;
   vp8_diamond_search_fn_t diamond_search_sad;
   vp8_variance_fn_ptr_t fn_ptr[BLOCK_MAX_SEGMENTS];
diff --git a/vp8/encoder/rdopt.c b/vp8/encoder/rdopt.c
index 79a858e43..5821fc734 100644
--- a/vp8/encoder/rdopt.c
+++ b/vp8/encoder/rdopt.c
@@ -1097,8 +1097,8 @@ static void rd_check_segment(VP8_COMP *cpi, MACROBLOCK *x, BEST_SEG_INFO *bsi,
             vp8_clamp_mv(&mvp_full, x->mv_col_min, x->mv_col_max, x->mv_row_min,
                          x->mv_row_max);
 
-            thissme = cpi->full_search_sad(x, c, e, &mvp_full, sadpb, 16,
-                                           v_fn_ptr, x->mvcost, bsi->ref_mv);
+            thissme = vp8_full_search_sad(x, c, e, &mvp_full, sadpb, 16,
+                                          v_fn_ptr, x->mvcost, bsi->ref_mv);
 
             if (thissme < bestsme) {
               bestsme = thissme;
diff --git a/vp9/encoder/vp9_encoder.c b/vp9/encoder/vp9_encoder.c
index 97805fc16..84ab80fe3 100644
--- a/vp9/encoder/vp9_encoder.c
+++ b/vp9/encoder/vp9_encoder.c
@@ -1569,15 +1569,13 @@ void vp9_set_rc_buffer_sizes(VP9_COMP *cpi) {
 }
 
 #if CONFIG_VP9_HIGHBITDEPTH
-// TODO(angiebird): make sdx8f available for highbitdepth if needed
 #define HIGHBD_BFP(BT, SDF, SDAF, VF, SVF, SVAF, SDX4DF) \
   cpi->fn_ptr[BT].sdf = SDF;                             \
   cpi->fn_ptr[BT].sdaf = SDAF;                           \
   cpi->fn_ptr[BT].vf = VF;                               \
   cpi->fn_ptr[BT].svf = SVF;                             \
   cpi->fn_ptr[BT].svaf = SVAF;                           \
-  cpi->fn_ptr[BT].sdx4df = SDX4DF;                       \
-  cpi->fn_ptr[BT].sdx8f = NULL;
+  cpi->fn_ptr[BT].sdx4df = SDX4DF;
 
 #define MAKE_BFP_SAD_WRAPPER(fnname)                                           \
   static unsigned int fnname##_bits8(const uint8_t *src_ptr,                   \
@@ -2561,67 +2559,61 @@ VP9_COMP *vp9_create_compressor(const VP9EncoderConfig *oxcf,
   CHECK_MEM_ERROR(cm, cpi->source_diff_var, vpx_calloc(cm->MBs, sizeof(diff)));
   cpi->source_var_thresh = 0;
   cpi->frames_till_next_var_check = 0;
-#define BFP(BT, SDF, SDAF, VF, SVF, SVAF, SDX4DF, SDX8F) \
-  cpi->fn_ptr[BT].sdf = SDF;                             \
-  cpi->fn_ptr[BT].sdaf = SDAF;                           \
-  cpi->fn_ptr[BT].vf = VF;                               \
-  cpi->fn_ptr[BT].svf = SVF;                             \
-  cpi->fn_ptr[BT].svaf = SVAF;                           \
-  cpi->fn_ptr[BT].sdx4df = SDX4DF;                       \
-  cpi->fn_ptr[BT].sdx8f = SDX8F;
+#define BFP(BT, SDF, SDAF, VF, SVF, SVAF, SDX4DF) \
+  cpi->fn_ptr[BT].sdf = SDF;                      \
+  cpi->fn_ptr[BT].sdaf = SDAF;                    \
+  cpi->fn_ptr[BT].vf = VF;                        \
+  cpi->fn_ptr[BT].svf = SVF;                      \
+  cpi->fn_ptr[BT].svaf = SVAF;                    \
+  cpi->fn_ptr[BT].sdx4df = SDX4DF;
 
-  // TODO(angiebird): make sdx8f available for every block size
   BFP(BLOCK_32X16, vpx_sad32x16, vpx_sad32x16_avg, vpx_variance32x16,
       vpx_sub_pixel_variance32x16, vpx_sub_pixel_avg_variance32x16,
-      vpx_sad32x16x4d, NULL)
+      vpx_sad32x16x4d)
 
   BFP(BLOCK_16X32, vpx_sad16x32, vpx_sad16x32_avg, vpx_variance16x32,
       vpx_sub_pixel_variance16x32, vpx_sub_pixel_avg_variance16x32,
-      vpx_sad16x32x4d, NULL)
+      vpx_sad16x32x4d)
 
   BFP(BLOCK_64X32, vpx_sad64x32, vpx_sad64x32_avg, vpx_variance64x32,
       vpx_sub_pixel_variance64x32, vpx_sub_pixel_avg_variance64x32,
-      vpx_sad64x32x4d, NULL)
+      vpx_sad64x32x4d)
 
   BFP(BLOCK_32X64, vpx_sad32x64, vpx_sad32x64_avg, vpx_variance32x64,
       vpx_sub_pixel_variance32x64, vpx_sub_pixel_avg_variance32x64,
-      vpx_sad32x64x4d, NULL)
+      vpx_sad32x64x4d)
 
   BFP(BLOCK_32X32, vpx_sad32x32, vpx_sad32x32_avg, vpx_variance32x32,
       vpx_sub_pixel_variance32x32, vpx_sub_pixel_avg_variance32x32,
-      vpx_sad32x32x4d, vpx_sad32x32x8)
+      vpx_sad32x32x4d)
 
   BFP(BLOCK_64X64, vpx_sad64x64, vpx_sad64x64_avg, vpx_variance64x64,
       vpx_sub_pixel_variance64x64, vpx_sub_pixel_avg_variance64x64,
-      vpx_sad64x64x4d, NULL)
+      vpx_sad64x64x4d)
 
   BFP(BLOCK_16X16, vpx_sad16x16, vpx_sad16x16_avg, vpx_variance16x16,
       vpx_sub_pixel_variance16x16, vpx_sub_pixel_avg_variance16x16,
-      vpx_sad16x16x4d, vpx_sad16x16x8)
+      vpx_sad16x16x4d)
 
   BFP(BLOCK_16X8, vpx_sad16x8, vpx_sad16x8_avg, vpx_variance16x8,
       vpx_sub_pixel_variance16x8, vpx_sub_pixel_avg_variance16x8,
-      vpx_sad16x8x4d, vpx_sad16x8x8)
+      vpx_sad16x8x4d)
 
   BFP(BLOCK_8X16, vpx_sad8x16, vpx_sad8x16_avg, vpx_variance8x16,
       vpx_sub_pixel_variance8x16, vpx_sub_pixel_avg_variance8x16,
-      vpx_sad8x16x4d, vpx_sad8x16x8)
+      vpx_sad8x16x4d)
 
   BFP(BLOCK_8X8, vpx_sad8x8, vpx_sad8x8_avg, vpx_variance8x8,
-      vpx_sub_pixel_variance8x8, vpx_sub_pixel_avg_variance8x8, vpx_sad8x8x4d,
-      vpx_sad8x8x8)
+      vpx_sub_pixel_variance8x8, vpx_sub_pixel_avg_variance8x8, vpx_sad8x8x4d)
 
   BFP(BLOCK_8X4, vpx_sad8x4, vpx_sad8x4_avg, vpx_variance8x4,
-      vpx_sub_pixel_variance8x4, vpx_sub_pixel_avg_variance8x4, vpx_sad8x4x4d,
-      NULL)
+      vpx_sub_pixel_variance8x4, vpx_sub_pixel_avg_variance8x4, vpx_sad8x4x4d)
 
   BFP(BLOCK_4X8, vpx_sad4x8, vpx_sad4x8_avg, vpx_variance4x8,
-      vpx_sub_pixel_variance4x8, vpx_sub_pixel_avg_variance4x8, vpx_sad4x8x4d,
-      NULL)
+      vpx_sub_pixel_variance4x8, vpx_sub_pixel_avg_variance4x8, vpx_sad4x8x4d)
 
   BFP(BLOCK_4X4, vpx_sad4x4, vpx_sad4x4_avg, vpx_variance4x4,
-      vpx_sub_pixel_variance4x4, vpx_sub_pixel_avg_variance4x4, vpx_sad4x4x4d,
-      vpx_sad4x4x8)
+      vpx_sub_pixel_variance4x4, vpx_sub_pixel_avg_variance4x4, vpx_sad4x4x4d)
 
 #if CONFIG_VP9_HIGHBITDEPTH
   highbd_set_var_fns(cpi);
diff --git a/vp9/encoder/vp9_mcomp.c b/vp9/encoder/vp9_mcomp.c
index cd6706420..1f08aa5de 100644
--- a/vp9/encoder/vp9_mcomp.c
+++ b/vp9/encoder/vp9_mcomp.c
@@ -1796,29 +1796,6 @@ static int64_t exhaustive_mesh_search_single_step(
   end_col = VPXMIN(center_mv->col + range, mv_limits->col_max);
   for (r = start_row; r <= end_row; r += 1) {
     c = start_col;
-    // sdx8f may not be available some block size
-    if (fn_ptr->sdx8f) {
-      while (c + 7 <= end_col) {
-        unsigned int sads[8];
-        const MV mv = { r, c };
-        const uint8_t *buf = get_buf_from_mv(pre, &mv);
-        fn_ptr->sdx8f(src->buf, src->stride, buf, pre->stride, sads);
-
-        for (i = 0; i < 8; ++i) {
-          int64_t sad = (int64_t)sads[i] << LOG2_PRECISION;
-          if (sad < best_sad) {
-            const MV mv = { r, c + i };
-            sad += lambda *
-                   vp9_nb_mvs_inconsistency(&mv, nb_full_mvs, full_mv_num);
-            if (sad < best_sad) {
-              best_sad = sad;
-              *best_mv = mv;
-            }
-          }
-        }
-        c += 8;
-      }
-    }
     while (c + 3 <= end_col) {
       unsigned int sads[4];
       const uint8_t *addrs[4];
diff --git a/vp9/encoder/vp9_mcomp.h b/vp9/encoder/vp9_mcomp.h
index 0c4d8f23c..bdaf2ce77 100644
--- a/vp9/encoder/vp9_mcomp.h
+++ b/vp9/encoder/vp9_mcomp.h
@@ -93,16 +93,6 @@ extern fractional_mv_step_fp vp9_skip_sub_pixel_tree;
 extern fractional_mv_step_fp vp9_return_max_sub_pixel_mv;
 extern fractional_mv_step_fp vp9_return_min_sub_pixel_mv;
 
-typedef int (*vp9_full_search_fn_t)(const MACROBLOCK *x, const MV *ref_mv,
-                                    int sad_per_bit, int distance,
-                                    const vp9_variance_fn_ptr_t *fn_ptr,
-                                    const MV *center_mv, MV *best_mv);
-
-typedef int (*vp9_refining_search_fn_t)(const MACROBLOCK *x, MV *ref_mv,
-                                        int sad_per_bit, int distance,
-                                        const vp9_variance_fn_ptr_t *fn_ptr,
-                                        const MV *center_mv);
-
 typedef int (*vp9_diamond_search_fn_t)(
     const MACROBLOCK *x, const search_site_config *cfg, MV *ref_mv, MV *best_mv,
     int search_param, int sad_per_bit, int *num00,
diff --git a/vpx_dsp/mips/sad_mmi.c b/vpx_dsp/mips/sad_mmi.c
index eaca4773f..7f5882bca 100644
--- a/vpx_dsp/mips/sad_mmi.c
+++ b/vpx_dsp/mips/sad_mmi.c
@@ -334,19 +334,6 @@
   "paddw      %[ftmp3],   %[ftmp3],       %[ftmp1]            \n\t"
 #endif /* _MIPS_SIM == _ABIO32 */
 
-// depending on call sites, pass **ref_array to avoid & in subsequent call and
-// de-dup with 4D below.
-#define sadMxNxK_mmi(m, n, k)                                                 \
-  void vpx_sad##m##x##n##x##k##_mmi(const uint8_t *src, int src_stride,       \
-                                    const uint8_t *ref_array, int ref_stride, \
-                                    uint32_t *sad_array) {                    \
-    int i;                                                                    \
-    for (i = 0; i < (k); ++i)                                                 \
-      sad_array[i] =                                                          \
-          vpx_sad##m##x##n##_mmi(src, src_stride, &ref_array[i], ref_stride); \
-  }
-
-// This appears to be equivalent to the above when k == 4 and refs is const
 #define sadMxNx4D_mmi(m, n)                                                  \
   void vpx_sad##m##x##n##x4d_mmi(const uint8_t *src, int src_stride,         \
                                  const uint8_t *const ref_array[],           \
@@ -583,10 +570,6 @@ static inline unsigned int vpx_sad16x(const uint8_t *src, int src_stride,
 vpx_sad16xN(32);
 vpx_sad16xN(16);
 vpx_sad16xN(8);
-sadMxNxK_mmi(16, 16, 3);
-sadMxNxK_mmi(16, 16, 8);
-sadMxNxK_mmi(16, 8, 3);
-sadMxNxK_mmi(16, 8, 8);
 sadMxNx4D_mmi(16, 32);
 sadMxNx4D_mmi(16, 16);
 sadMxNx4D_mmi(16, 8);
@@ -681,10 +664,6 @@ static inline unsigned int vpx_sad8x(const uint8_t *src, int src_stride,
 vpx_sad8xN(16);
 vpx_sad8xN(8);
 vpx_sad8xN(4);
-sadMxNxK_mmi(8, 16, 3);
-sadMxNxK_mmi(8, 16, 8);
-sadMxNxK_mmi(8, 8, 3);
-sadMxNxK_mmi(8, 8, 8);
 sadMxNx4D_mmi(8, 16);
 sadMxNx4D_mmi(8, 8);
 sadMxNx4D_mmi(8, 4);
@@ -777,8 +756,6 @@ static inline unsigned int vpx_sad4x(const uint8_t *src, int src_stride,
 
 vpx_sad4xN(8);
 vpx_sad4xN(4);
-sadMxNxK_mmi(4, 4, 3);
-sadMxNxK_mmi(4, 4, 8);
 sadMxNx4D_mmi(4, 8);
 sadMxNx4D_mmi(4, 4);
 
diff --git a/vpx_dsp/mips/sad_msa.c b/vpx_dsp/mips/sad_msa.c
index e3e91c433..b0f8ff1fd 100644
--- a/vpx_dsp/mips/sad_msa.c
+++ b/vpx_dsp/mips/sad_msa.c
@@ -159,380 +159,6 @@ static uint32_t sad_64width_msa(const uint8_t *src, int32_t src_stride,
   return sad;
 }
 
-static void sad_4width_x3_msa(const uint8_t *src_ptr, int32_t src_stride,
-                              const uint8_t *ref_ptr, int32_t ref_stride,
-                              int32_t height, uint32_t *sad_array) {
-  int32_t ht_cnt;
-  uint32_t src0, src1, src2, src3;
-  v16u8 src = { 0 };
-  v16u8 ref = { 0 };
-  v16u8 ref0, ref1, ref2, ref3, diff;
-  v8u16 sad0 = { 0 };
-  v8u16 sad1 = { 0 };
-  v8u16 sad2 = { 0 };
-
-  for (ht_cnt = (height >> 2); ht_cnt--;) {
-    LW4(src_ptr, src_stride, src0, src1, src2, src3);
-    src_ptr += (4 * src_stride);
-    INSERT_W4_UB(src0, src1, src2, src3, src);
-
-    LD_UB4(ref_ptr, ref_stride, ref0, ref1, ref2, ref3);
-    ref_ptr += (4 * ref_stride);
-    SAD_INSVE_W4_UB(ref0, ref1, ref2, ref3, ref);
-    diff = __msa_asub_u_b(src, ref);
-    sad0 += __msa_hadd_u_h(diff, diff);
-
-    SLDI_B2_UB(ref0, ref1, ref0, ref1, ref0, ref1, 1);
-    SLDI_B2_UB(ref2, ref3, ref2, ref3, ref2, ref3, 1);
-    SAD_INSVE_W4_UB(ref0, ref1, ref2, ref3, ref);
-    diff = __msa_asub_u_b(src, ref);
-    sad1 += __msa_hadd_u_h(diff, diff);
-
-    SLDI_B2_UB(ref0, ref1, ref0, ref1, ref0, ref1, 1);
-    SLDI_B2_UB(ref2, ref3, ref2, ref3, ref2, ref3, 1);
-    SAD_INSVE_W4_UB(ref0, ref1, ref2, ref3, ref);
-    diff = __msa_asub_u_b(src, ref);
-    sad2 += __msa_hadd_u_h(diff, diff);
-  }
-
-  sad_array[0] = HADD_UH_U32(sad0);
-  sad_array[1] = HADD_UH_U32(sad1);
-  sad_array[2] = HADD_UH_U32(sad2);
-}
-
-static void sad_8width_x3_msa(const uint8_t *src, int32_t src_stride,
-                              const uint8_t *ref, int32_t ref_stride,
-                              int32_t height, uint32_t *sad_array) {
-  int32_t ht_cnt;
-  v16u8 src0, src1, src2, src3;
-  v16u8 ref0, ref1, ref00, ref11, ref22, ref33;
-  v8u16 sad0 = { 0 };
-  v8u16 sad1 = { 0 };
-  v8u16 sad2 = { 0 };
-
-  for (ht_cnt = (height >> 2); ht_cnt--;) {
-    LD_UB4(src, src_stride, src0, src1, src2, src3);
-    src += (4 * src_stride);
-    LD_UB4(ref, ref_stride, ref00, ref11, ref22, ref33);
-    ref += (4 * ref_stride);
-    PCKEV_D4_UB(src1, src0, src3, src2, ref11, ref00, ref33, ref22, src0, src1,
-                ref0, ref1);
-    sad0 += SAD_UB2_UH(src0, src1, ref0, ref1);
-
-    SLDI_B2_UB(ref00, ref11, ref00, ref11, ref00, ref11, 1);
-    SLDI_B2_UB(ref22, ref33, ref22, ref33, ref22, ref33, 1);
-    PCKEV_D2_UB(ref11, ref00, ref33, ref22, ref0, ref1);
-    sad1 += SAD_UB2_UH(src0, src1, ref0, ref1);
-
-    SLDI_B2_UB(ref00, ref11, ref00, ref11, ref00, ref11, 1);
-    SLDI_B2_UB(ref22, ref33, ref22, ref33, ref22, ref33, 1);
-    PCKEV_D2_UB(ref11, ref00, ref33, ref22, ref0, ref1);
-    sad2 += SAD_UB2_UH(src0, src1, ref0, ref1);
-  }
-
-  sad_array[0] = HADD_UH_U32(sad0);
-  sad_array[1] = HADD_UH_U32(sad1);
-  sad_array[2] = HADD_UH_U32(sad2);
-}
-
-static void sad_16width_x3_msa(const uint8_t *src_ptr, int32_t src_stride,
-                               const uint8_t *ref_ptr, int32_t ref_stride,
-                               int32_t height, uint32_t *sad_array) {
-  int32_t ht_cnt;
-  v16u8 src, ref, ref0, ref1, diff;
-  v8u16 sad0 = { 0 };
-  v8u16 sad1 = { 0 };
-  v8u16 sad2 = { 0 };
-
-  for (ht_cnt = (height >> 1); ht_cnt--;) {
-    src = LD_UB(src_ptr);
-    src_ptr += src_stride;
-    LD_UB2(ref_ptr, 16, ref0, ref1);
-    ref_ptr += ref_stride;
-
-    diff = __msa_asub_u_b(src, ref0);
-    sad0 += __msa_hadd_u_h(diff, diff);
-
-    ref = (v16u8)__msa_sldi_b((v16i8)ref1, (v16i8)ref0, 1);
-    diff = __msa_asub_u_b(src, ref);
-    sad1 += __msa_hadd_u_h(diff, diff);
-
-    ref = (v16u8)__msa_sldi_b((v16i8)ref1, (v16i8)ref0, 2);
-    diff = __msa_asub_u_b(src, ref);
-    sad2 += __msa_hadd_u_h(diff, diff);
-
-    src = LD_UB(src_ptr);
-    src_ptr += src_stride;
-    LD_UB2(ref_ptr, 16, ref0, ref1);
-    ref_ptr += ref_stride;
-
-    diff = __msa_asub_u_b(src, ref0);
-    sad0 += __msa_hadd_u_h(diff, diff);
-
-    ref = (v16u8)__msa_sldi_b((v16i8)ref1, (v16i8)ref0, 1);
-    diff = __msa_asub_u_b(src, ref);
-    sad1 += __msa_hadd_u_h(diff, diff);
-
-    ref = (v16u8)__msa_sldi_b((v16i8)ref1, (v16i8)ref0, 2);
-    diff = __msa_asub_u_b(src, ref);
-    sad2 += __msa_hadd_u_h(diff, diff);
-  }
-
-  sad_array[0] = HADD_UH_U32(sad0);
-  sad_array[1] = HADD_UH_U32(sad1);
-  sad_array[2] = HADD_UH_U32(sad2);
-}
-
-static void sad_4width_x8_msa(const uint8_t *src_ptr, int32_t src_stride,
-                              const uint8_t *ref_ptr, int32_t ref_stride,
-                              int32_t height, uint32_t *sad_array) {
-  int32_t ht_cnt;
-  uint32_t src0, src1, src2, src3;
-  v16u8 ref0, ref1, ref2, ref3, diff;
-  v16u8 src = { 0 };
-  v16u8 ref = { 0 };
-  v8u16 sad0 = { 0 };
-  v8u16 sad1 = { 0 };
-  v8u16 sad2 = { 0 };
-  v8u16 sad3 = { 0 };
-  v8u16 sad4 = { 0 };
-  v8u16 sad5 = { 0 };
-  v8u16 sad6 = { 0 };
-  v8u16 sad7 = { 0 };
-
-  for (ht_cnt = (height >> 2); ht_cnt--;) {
-    LW4(src_ptr, src_stride, src0, src1, src2, src3);
-    INSERT_W4_UB(src0, src1, src2, src3, src);
-    src_ptr += (4 * src_stride);
-    LD_UB4(ref_ptr, ref_stride, ref0, ref1, ref2, ref3);
-    ref_ptr += (4 * ref_stride);
-
-    SAD_INSVE_W4_UB(ref0, ref1, ref2, ref3, ref);
-    diff = __msa_asub_u_b(src, ref);
-    sad0 += __msa_hadd_u_h(diff, diff);
-
-    SLDI_B2_UB(ref0, ref1, ref0, ref1, ref0, ref1, 1);
-    SLDI_B2_UB(ref2, ref3, ref2, ref3, ref2, ref3, 1);
-    SAD_INSVE_W4_UB(ref0, ref1, ref2, ref3, ref);
-    diff = __msa_asub_u_b(src, ref);
-    sad1 += __msa_hadd_u_h(diff, diff);
-
-    SLDI_B2_UB(ref0, ref1, ref0, ref1, ref0, ref1, 1);
-    SLDI_B2_UB(ref2, ref3, ref2, ref3, ref2, ref3, 1);
-    SAD_INSVE_W4_UB(ref0, ref1, ref2, ref3, ref);
-    diff = __msa_asub_u_b(src, ref);
-    sad2 += __msa_hadd_u_h(diff, diff);
-
-    SLDI_B2_UB(ref0, ref1, ref0, ref1, ref0, ref1, 1);
-    SLDI_B2_UB(ref2, ref3, ref2, ref3, ref2, ref3, 1);
-    SAD_INSVE_W4_UB(ref0, ref1, ref2, ref3, ref);
-    diff = __msa_asub_u_b(src, ref);
-    sad3 += __msa_hadd_u_h(diff, diff);
-
-    SLDI_B2_UB(ref0, ref1, ref0, ref1, ref0, ref1, 1);
-    SLDI_B2_UB(ref2, ref3, ref2, ref3, ref2, ref3, 1);
-    SAD_INSVE_W4_UB(ref0, ref1, ref2, ref3, ref);
-    diff = __msa_asub_u_b(src, ref);
-    sad4 += __msa_hadd_u_h(diff, diff);
-
-    SLDI_B2_UB(ref0, ref1, ref0, ref1, ref0, ref1, 1);
-    SLDI_B2_UB(ref2, ref3, ref2, ref3, ref2, ref3, 1);
-    SAD_INSVE_W4_UB(ref0, ref1, ref2, ref3, ref);
-    diff = __msa_asub_u_b(src, ref);
-    sad5 += __msa_hadd_u_h(diff, diff);
-
-    SLDI_B2_UB(ref0, ref1, ref0, ref1, ref0, ref1, 1);
-    SLDI_B2_UB(ref2, ref3, ref2, ref3, ref2, ref3, 1);
-    SAD_INSVE_W4_UB(ref0, ref1, ref2, ref3, ref);
-    diff = __msa_asub_u_b(src, ref);
-    sad6 += __msa_hadd_u_h(diff, diff);
-
-    SLDI_B2_UB(ref0, ref1, ref0, ref1, ref0, ref1, 1);
-    SLDI_B2_UB(ref2, ref3, ref2, ref3, ref2, ref3, 1);
-    SAD_INSVE_W4_UB(ref0, ref1, ref2, ref3, ref);
-    diff = __msa_asub_u_b(src, ref);
-    sad7 += __msa_hadd_u_h(diff, diff);
-  }
-
-  sad_array[0] = HADD_UH_U32(sad0);
-  sad_array[1] = HADD_UH_U32(sad1);
-  sad_array[2] = HADD_UH_U32(sad2);
-  sad_array[3] = HADD_UH_U32(sad3);
-  sad_array[4] = HADD_UH_U32(sad4);
-  sad_array[5] = HADD_UH_U32(sad5);
-  sad_array[6] = HADD_UH_U32(sad6);
-  sad_array[7] = HADD_UH_U32(sad7);
-}
-
-static void sad_8width_x8_msa(const uint8_t *src, int32_t src_stride,
-                              const uint8_t *ref, int32_t ref_stride,
-                              int32_t height, uint32_t *sad_array) {
-  int32_t ht_cnt;
-  v16u8 src0, src1, src2, src3;
-  v16u8 ref0, ref1, ref00, ref11, ref22, ref33;
-  v8u16 sad0 = { 0 };
-  v8u16 sad1 = { 0 };
-  v8u16 sad2 = { 0 };
-  v8u16 sad3 = { 0 };
-  v8u16 sad4 = { 0 };
-  v8u16 sad5 = { 0 };
-  v8u16 sad6 = { 0 };
-  v8u16 sad7 = { 0 };
-
-  for (ht_cnt = (height >> 2); ht_cnt--;) {
-    LD_UB4(src, src_stride, src0, src1, src2, src3);
-    src += (4 * src_stride);
-    LD_UB4(ref, ref_stride, ref00, ref11, ref22, ref33);
-    ref += (4 * ref_stride);
-    PCKEV_D4_UB(src1, src0, src3, src2, ref11, ref00, ref33, ref22, src0, src1,
-                ref0, ref1);
-    sad0 += SAD_UB2_UH(src0, src1, ref0, ref1);
-
-    SLDI_B2_UB(ref00, ref11, ref00, ref11, ref00, ref11, 1);
-    SLDI_B2_UB(ref22, ref33, ref22, ref33, ref22, ref33, 1);
-    PCKEV_D2_UB(ref11, ref00, ref33, ref22, ref0, ref1);
-    sad1 += SAD_UB2_UH(src0, src1, ref0, ref1);
-
-    SLDI_B2_UB(ref00, ref11, ref00, ref11, ref00, ref11, 1);
-    SLDI_B2_UB(ref22, ref33, ref22, ref33, ref22, ref33, 1);
-    PCKEV_D2_UB(ref11, ref00, ref33, ref22, ref0, ref1);
-    sad2 += SAD_UB2_UH(src0, src1, ref0, ref1);
-
-    SLDI_B2_UB(ref00, ref11, ref00, ref11, ref00, ref11, 1);
-    SLDI_B2_UB(ref22, ref33, ref22, ref33, ref22, ref33, 1);
-    PCKEV_D2_UB(ref11, ref00, ref33, ref22, ref0, ref1);
-    sad3 += SAD_UB2_UH(src0, src1, ref0, ref1);
-
-    SLDI_B2_UB(ref00, ref11, ref00, ref11, ref00, ref11, 1);
-    SLDI_B2_UB(ref22, ref33, ref22, ref33, ref22, ref33, 1);
-    PCKEV_D2_UB(ref11, ref00, ref33, ref22, ref0, ref1);
-    sad4 += SAD_UB2_UH(src0, src1, ref0, ref1);
-
-    SLDI_B2_UB(ref00, ref11, ref00, ref11, ref00, ref11, 1);
-    SLDI_B2_UB(ref22, ref33, ref22, ref33, ref22, ref33, 1);
-    PCKEV_D2_UB(ref11, ref00, ref33, ref22, ref0, ref1);
-    sad5 += SAD_UB2_UH(src0, src1, ref0, ref1);
-
-    SLDI_B2_UB(ref00, ref11, ref00, ref11, ref00, ref11, 1);
-    SLDI_B2_UB(ref22, ref33, ref22, ref33, ref22, ref33, 1);
-    PCKEV_D2_UB(ref11, ref00, ref33, ref22, ref0, ref1);
-    sad6 += SAD_UB2_UH(src0, src1, ref0, ref1);
-
-    SLDI_B2_UB(ref00, ref11, ref00, ref11, ref00, ref11, 1);
-    SLDI_B2_UB(ref22, ref33, ref22, ref33, ref22, ref33, 1);
-    PCKEV_D2_UB(ref11, ref00, ref33, ref22, ref0, ref1);
-    sad7 += SAD_UB2_UH(src0, src1, ref0, ref1);
-  }
-
-  sad_array[0] = HADD_UH_U32(sad0);
-  sad_array[1] = HADD_UH_U32(sad1);
-  sad_array[2] = HADD_UH_U32(sad2);
-  sad_array[3] = HADD_UH_U32(sad3);
-  sad_array[4] = HADD_UH_U32(sad4);
-  sad_array[5] = HADD_UH_U32(sad5);
-  sad_array[6] = HADD_UH_U32(sad6);
-  sad_array[7] = HADD_UH_U32(sad7);
-}
-
-static void sad_16width_x8_msa(const uint8_t *src_ptr, int32_t src_stride,
-                               const uint8_t *ref_ptr, int32_t ref_stride,
-                               int32_t height, uint32_t *sad_array) {
-  int32_t ht_cnt;
-  v16u8 src, ref0, ref1, ref;
-  v16u8 diff;
-  v8u16 sad0 = { 0 };
-  v8u16 sad1 = { 0 };
-  v8u16 sad2 = { 0 };
-  v8u16 sad3 = { 0 };
-  v8u16 sad4 = { 0 };
-  v8u16 sad5 = { 0 };
-  v8u16 sad6 = { 0 };
-  v8u16 sad7 = { 0 };
-
-  for (ht_cnt = (height >> 1); ht_cnt--;) {
-    src = LD_UB(src_ptr);
-    src_ptr += src_stride;
-    LD_UB2(ref_ptr, 16, ref0, ref1);
-    ref_ptr += ref_stride;
-
-    diff = __msa_asub_u_b(src, ref0);
-    sad0 += __msa_hadd_u_h(diff, diff);
-
-    ref = (v16u8)__msa_sldi_b((v16i8)ref1, (v16i8)ref0, 1);
-    diff = __msa_asub_u_b(src, ref);
-    sad1 += __msa_hadd_u_h(diff, diff);
-
-    ref = (v16u8)__msa_sldi_b((v16i8)ref1, (v16i8)ref0, 2);
-    diff = __msa_asub_u_b(src, ref);
-    sad2 += __msa_hadd_u_h(diff, diff);
-
-    ref = (v16u8)__msa_sldi_b((v16i8)ref1, (v16i8)ref0, 3);
-    diff = __msa_asub_u_b(src, ref);
-    sad3 += __msa_hadd_u_h(diff, diff);
-
-    ref = (v16u8)__msa_sldi_b((v16i8)ref1, (v16i8)ref0, 4);
-    diff = __msa_asub_u_b(src, ref);
-    sad4 += __msa_hadd_u_h(diff, diff);
-
-    ref = (v16u8)__msa_sldi_b((v16i8)ref1, (v16i8)ref0, 5);
-    diff = __msa_asub_u_b(src, ref);
-    sad5 += __msa_hadd_u_h(diff, diff);
-
-    ref = (v16u8)__msa_sldi_b((v16i8)ref1, (v16i8)ref0, 6);
-    diff = __msa_asub_u_b(src, ref);
-    sad6 += __msa_hadd_u_h(diff, diff);
-
-    ref = (v16u8)__msa_sldi_b((v16i8)ref1, (v16i8)ref0, 7);
-    diff = __msa_asub_u_b(src, ref);
-    sad7 += __msa_hadd_u_h(diff, diff);
-
-    src = LD_UB(src_ptr);
-    src_ptr += src_stride;
-    LD_UB2(ref_ptr, 16, ref0, ref1);
-    ref_ptr += ref_stride;
-
-    diff = __msa_asub_u_b(src, ref0);
-    sad0 += __msa_hadd_u_h(diff, diff);
-
-    ref = (v16u8)__msa_sldi_b((v16i8)ref1, (v16i8)ref0, 1);
-    diff = __msa_asub_u_b(src, ref);
-    sad1 += __msa_hadd_u_h(diff, diff);
-
-    ref = (v16u8)__msa_sldi_b((v16i8)ref1, (v16i8)ref0, 2);
-    diff = __msa_asub_u_b(src, ref);
-    sad2 += __msa_hadd_u_h(diff, diff);
-
-    ref = (v16u8)__msa_sldi_b((v16i8)ref1, (v16i8)ref0, 3);
-    diff = __msa_asub_u_b(src, ref);
-    sad3 += __msa_hadd_u_h(diff, diff);
-
-    ref = (v16u8)__msa_sldi_b((v16i8)ref1, (v16i8)ref0, 4);
-    diff = __msa_asub_u_b(src, ref);
-    sad4 += __msa_hadd_u_h(diff, diff);
-
-    ref = (v16u8)__msa_sldi_b((v16i8)ref1, (v16i8)ref0, 5);
-    diff = __msa_asub_u_b(src, ref);
-    sad5 += __msa_hadd_u_h(diff, diff);
-
-    ref = (v16u8)__msa_sldi_b((v16i8)ref1, (v16i8)ref0, 6);
-    diff = __msa_asub_u_b(src, ref);
-    sad6 += __msa_hadd_u_h(diff, diff);
-
-    ref = (v16u8)__msa_sldi_b((v16i8)ref1, (v16i8)ref0, 7);
-    diff = __msa_asub_u_b(src, ref);
-    sad7 += __msa_hadd_u_h(diff, diff);
-  }
-
-  sad_array[0] = HADD_UH_U32(sad0);
-  sad_array[1] = HADD_UH_U32(sad1);
-  sad_array[2] = HADD_UH_U32(sad2);
-  sad_array[3] = HADD_UH_U32(sad3);
-  sad_array[4] = HADD_UH_U32(sad4);
-  sad_array[5] = HADD_UH_U32(sad5);
-  sad_array[6] = HADD_UH_U32(sad6);
-  sad_array[7] = HADD_UH_U32(sad7);
-}
-
 static void sad_4width_x4d_msa(const uint8_t *src_ptr, int32_t src_stride,
                                const uint8_t *const aref_ptr[],
                                int32_t ref_stride, int32_t height,
@@ -1037,48 +663,6 @@ static uint32_t avgsad_64width_msa(const uint8_t *src, int32_t src_stride,
     return sad_64width_msa(src, src_stride, ref, ref_stride, height);         \
   }
 
-#define VPX_SAD_4xHEIGHTx3_MSA(height)                                   \
-  void vpx_sad4x##height##x3_msa(const uint8_t *src, int32_t src_stride, \
-                                 const uint8_t *ref, int32_t ref_stride, \
-                                 uint32_t sads[3]) {                     \
-    sad_4width_x3_msa(src, src_stride, ref, ref_stride, height, sads);   \
-  }
-
-#define VPX_SAD_8xHEIGHTx3_MSA(height)                                   \
-  void vpx_sad8x##height##x3_msa(const uint8_t *src, int32_t src_stride, \
-                                 const uint8_t *ref, int32_t ref_stride, \
-                                 uint32_t sads[3]) {                     \
-    sad_8width_x3_msa(src, src_stride, ref, ref_stride, height, sads);   \
-  }
-
-#define VPX_SAD_16xHEIGHTx3_MSA(height)                                   \
-  void vpx_sad16x##height##x3_msa(const uint8_t *src, int32_t src_stride, \
-                                  const uint8_t *ref, int32_t ref_stride, \
-                                  uint32_t sads[3]) {                     \
-    sad_16width_x3_msa(src, src_stride, ref, ref_stride, height, sads);   \
-  }
-
-#define VPX_SAD_4xHEIGHTx8_MSA(height)                                   \
-  void vpx_sad4x##height##x8_msa(const uint8_t *src, int32_t src_stride, \
-                                 const uint8_t *ref, int32_t ref_stride, \
-                                 uint32_t sads[8]) {                     \
-    sad_4width_x8_msa(src, src_stride, ref, ref_stride, height, sads);   \
-  }
-
-#define VPX_SAD_8xHEIGHTx8_MSA(height)                                   \
-  void vpx_sad8x##height##x8_msa(const uint8_t *src, int32_t src_stride, \
-                                 const uint8_t *ref, int32_t ref_stride, \
-                                 uint32_t sads[8]) {                     \
-    sad_8width_x8_msa(src, src_stride, ref, ref_stride, height, sads);   \
-  }
-
-#define VPX_SAD_16xHEIGHTx8_MSA(height)                                   \
-  void vpx_sad16x##height##x8_msa(const uint8_t *src, int32_t src_stride, \
-                                  const uint8_t *ref, int32_t ref_stride, \
-                                  uint32_t sads[8]) {                     \
-    sad_16width_x8_msa(src, src_stride, ref, ref_stride, height, sads);   \
-  }
-
 #define VPX_SAD_4xHEIGHTx4D_MSA(height)                                   \
   void vpx_sad4x##height##x4d_msa(const uint8_t *src, int32_t src_stride, \
                                   const uint8_t *const refs[4],           \
@@ -1186,29 +770,21 @@ VPX_AVGSAD_16xHEIGHT_MSA(32);
 
 // 16x16
 VPX_SAD_16xHEIGHT_MSA(16);
-VPX_SAD_16xHEIGHTx3_MSA(16);
-VPX_SAD_16xHEIGHTx8_MSA(16);
 VPX_SAD_16xHEIGHTx4D_MSA(16);
 VPX_AVGSAD_16xHEIGHT_MSA(16);
 
 // 16x8
 VPX_SAD_16xHEIGHT_MSA(8);
-VPX_SAD_16xHEIGHTx3_MSA(8);
-VPX_SAD_16xHEIGHTx8_MSA(8);
 VPX_SAD_16xHEIGHTx4D_MSA(8);
 VPX_AVGSAD_16xHEIGHT_MSA(8);
 
 // 8x16
 VPX_SAD_8xHEIGHT_MSA(16);
-VPX_SAD_8xHEIGHTx3_MSA(16);
-VPX_SAD_8xHEIGHTx8_MSA(16);
 VPX_SAD_8xHEIGHTx4D_MSA(16);
 VPX_AVGSAD_8xHEIGHT_MSA(16);
 
 // 8x8
 VPX_SAD_8xHEIGHT_MSA(8);
-VPX_SAD_8xHEIGHTx3_MSA(8);
-VPX_SAD_8xHEIGHTx8_MSA(8);
 VPX_SAD_8xHEIGHTx4D_MSA(8);
 VPX_AVGSAD_8xHEIGHT_MSA(8);
 
@@ -1224,7 +800,5 @@ VPX_AVGSAD_4xHEIGHT_MSA(8);
 
 // 4x4
 VPX_SAD_4xHEIGHT_MSA(4);
-VPX_SAD_4xHEIGHTx3_MSA(4);
-VPX_SAD_4xHEIGHTx8_MSA(4);
 VPX_SAD_4xHEIGHTx4D_MSA(4);
 VPX_AVGSAD_4xHEIGHT_MSA(4);
diff --git a/vpx_dsp/sad.c b/vpx_dsp/sad.c
index 46d513b68..b47c43430 100644
--- a/vpx_dsp/sad.c
+++ b/vpx_dsp/sad.c
@@ -45,35 +45,7 @@ static INLINE unsigned int sad(const uint8_t *src_ptr, int src_stride,
     return sad(src_ptr, src_stride, comp_pred, m, m, n);                      \
   }
 
-// Compare |src_ptr| to |k| adjacent blocks starting at |ref_ptr|.
-// |k| == {3,8}. Used in vp8 for an exhaustive search.
-// src:           ref:
-//  0  1  2  3     0  1  2  3  x  x
-//  4  5  6  7     6  7  8  9  x  x
-//  8  9 10 11    12 13 14 15  x  x
-// 12 13 14 15    18 19 20 21  x  x
-//
-//                 x  1  2  3  4  x
-//                 x  7  8  9 10  x
-//                 x 13 14 15 16  x
-//                 x 19 20 21 22  x
-//
-//                 x  x  2  3  4  5
-//                 x  x  8  9 10 11
-//                 x  x 14 15 16 17
-//                 x  x 20 21 22 23
-//
-#define sadMxNxK(m, n, k)                                                     \
-  void vpx_sad##m##x##n##x##k##_c(const uint8_t *src_ptr, int src_stride,     \
-                                  const uint8_t *ref_ptr, int ref_stride,     \
-                                  uint32_t sad_array[k]) {                    \
-    int i;                                                                    \
-    for (i = 0; i < k; ++i)                                                   \
-      sad_array[i] =                                                          \
-          vpx_sad##m##x##n##_c(src_ptr, src_stride, ref_ptr + i, ref_stride); \
-  }
-
-// Compare |src_ptr| to 4 distinct references in |ref_array[]|
+// Compare |src_ptr| to 4 distinct references in |ref_array[4]|
 #define sadMxNx4D(m, n)                                                        \
   void vpx_sad##m##x##n##x4d_c(const uint8_t *src_ptr, int src_stride,         \
                                const uint8_t *const ref_array[4],              \
@@ -99,7 +71,6 @@ sadMxNx4D(32, 64)
 
 // 32x32
 sadMxN(32, 32)
-sadMxNxK(32, 32, 8)
 sadMxNx4D(32, 32)
 
 // 32x16
@@ -112,26 +83,18 @@ sadMxNx4D(16, 32)
 
 // 16x16
 sadMxN(16, 16)
-sadMxNxK(16, 16, 3)
-sadMxNxK(16, 16, 8)
 sadMxNx4D(16, 16)
 
 // 16x8
 sadMxN(16, 8)
-sadMxNxK(16, 8, 3)
-sadMxNxK(16, 8, 8)
 sadMxNx4D(16, 8)
 
 // 8x16
 sadMxN(8, 16)
-sadMxNxK(8, 16, 3)
-sadMxNxK(8, 16, 8)
 sadMxNx4D(8, 16)
 
 // 8x8
 sadMxN(8, 8)
-sadMxNxK(8, 8, 3)
-sadMxNxK(8, 8, 8)
 sadMxNx4D(8, 8)
 
 // 8x4
@@ -144,8 +107,6 @@ sadMxNx4D(4, 8)
 
 // 4x4
 sadMxN(4, 4)
-sadMxNxK(4, 4, 3)
-sadMxNxK(4, 4, 8)
 sadMxNx4D(4, 4)
 /* clang-format on */
 
diff --git a/vpx_dsp/variance.h b/vpx_dsp/variance.h
index f8b44f03d..755cb907d 100644
--- a/vpx_dsp/variance.h
+++ b/vpx_dsp/variance.h
@@ -59,8 +59,6 @@ typedef struct variance_vtable {
   vpx_sad_fn_t sdf;
   vpx_variance_fn_t vf;
   vpx_subpixvariance_fn_t svf;
-  vpx_sad_multi_fn_t sdx3f;
-  vpx_sad_multi_fn_t sdx8f;
   vpx_sad_multi_d_fn_t sdx4df;
 #if VPX_ARCH_X86 || VPX_ARCH_X86_64
   vp8_copy32xn_fn_t copymem;
@@ -76,7 +74,6 @@ typedef struct vp9_variance_vtable {
   vpx_subpixvariance_fn_t svf;
   vpx_subp_avg_variance_fn_t svaf;
   vpx_sad_multi_d_fn_t sdx4df;
-  vpx_sad_multi_fn_t sdx8f;
 } vp9_variance_fn_ptr_t;
 #endif  // CONFIG_VP9
 
diff --git a/vpx_dsp/vpx_dsp.mk b/vpx_dsp/vpx_dsp.mk
index b930fbd0a..bf348c112 100644
--- a/vpx_dsp/vpx_dsp.mk
+++ b/vpx_dsp/vpx_dsp.mk
@@ -355,9 +355,6 @@ DSP_SRCS-$(HAVE_MSA)    += mips/subtract_msa.c
 DSP_SRCS-$(HAVE_MMI)    += mips/sad_mmi.c
 DSP_SRCS-$(HAVE_MMI)    += mips/subtract_mmi.c
 
-DSP_SRCS-$(HAVE_SSE3)   += x86/sad_sse3.asm
-DSP_SRCS-$(HAVE_SSSE3)  += x86/sad_ssse3.asm
-DSP_SRCS-$(HAVE_SSE4_1) += x86/sad_sse4.asm
 DSP_SRCS-$(HAVE_AVX2)   += x86/sad4d_avx2.c
 DSP_SRCS-$(HAVE_AVX2)   += x86/sad_avx2.c
 DSP_SRCS-$(HAVE_AVX512) += x86/sad4d_avx512.c
diff --git a/vpx_dsp/vpx_dsp_rtcd_defs.pl b/vpx_dsp/vpx_dsp_rtcd_defs.pl
index 63097b0b6..73f28ff92 100644
--- a/vpx_dsp/vpx_dsp_rtcd_defs.pl
+++ b/vpx_dsp/vpx_dsp_rtcd_defs.pl
@@ -866,44 +866,6 @@ add_proto qw/unsigned int vpx_sad4x4_avg/, "const uint8_t *src_ptr, int src_stri
 specialize qw/vpx_sad4x4_avg neon msa sse2 mmi/;
 
 #
-# Multi-block SAD, comparing a reference to N blocks 1 pixel apart horizontally
-#
-# Blocks of 3
-add_proto qw/void vpx_sad16x16x3/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t sad_array[3]";
-specialize qw/vpx_sad16x16x3 sse3 ssse3 msa mmi/;
-
-add_proto qw/void vpx_sad16x8x3/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t sad_array[3]";
-specialize qw/vpx_sad16x8x3 sse3 ssse3 msa mmi/;
-
-add_proto qw/void vpx_sad8x16x3/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t sad_array[3]";
-specialize qw/vpx_sad8x16x3 sse3 msa mmi/;
-
-add_proto qw/void vpx_sad8x8x3/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t sad_array[3]";
-specialize qw/vpx_sad8x8x3 sse3 msa mmi/;
-
-add_proto qw/void vpx_sad4x4x3/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t sad_array[3]";
-specialize qw/vpx_sad4x4x3 sse3 msa mmi/;
-
-# Blocks of 8
-add_proto qw/void vpx_sad32x32x8/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t sad_array[8]";
-specialize qw/vpx_sad32x32x8 avx2/;
-
-add_proto qw/void vpx_sad16x16x8/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t sad_array[8]";
-specialize qw/vpx_sad16x16x8 sse4_1 msa mmi/;
-
-add_proto qw/void vpx_sad16x8x8/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t sad_array[8]";
-specialize qw/vpx_sad16x8x8 sse4_1 msa mmi/;
-
-add_proto qw/void vpx_sad8x16x8/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t sad_array[8]";
-specialize qw/vpx_sad8x16x8 sse4_1 msa mmi/;
-
-add_proto qw/void vpx_sad8x8x8/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t sad_array[8]";
-specialize qw/vpx_sad8x8x8 sse4_1 msa mmi/;
-
-add_proto qw/void vpx_sad4x4x8/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t sad_array[8]";
-specialize qw/vpx_sad4x4x8 sse4_1 msa mmi/;
-
-#
 # Multi-block SAD, comparing a reference to N independent blocks
 #
 add_proto qw/void vpx_sad64x64x4d/, "const uint8_t *src_ptr, int src_stride, const uint8_t * const ref_array[4], int ref_stride, uint32_t sad_array[4]";
diff --git a/vpx_dsp/x86/sad4d_avx2.c b/vpx_dsp/x86/sad4d_avx2.c
index 81f1a916f..399b67b3f 100644
--- a/vpx_dsp/x86/sad4d_avx2.c
+++ b/vpx_dsp/x86/sad4d_avx2.c
@@ -73,63 +73,6 @@ void vpx_sad32x32x4d_avx2(const uint8_t *src_ptr, int src_stride,
   calc_final_4(sums, sad_array);
 }
 
-void vpx_sad32x32x8_avx2(const uint8_t *src_ptr, int src_stride,
-                         const uint8_t *ref_ptr, int ref_stride,
-                         uint32_t sad_array[8]) {
-  int i;
-  __m256i sums[8];
-
-  sums[0] = _mm256_setzero_si256();
-  sums[1] = _mm256_setzero_si256();
-  sums[2] = _mm256_setzero_si256();
-  sums[3] = _mm256_setzero_si256();
-  sums[4] = _mm256_setzero_si256();
-  sums[5] = _mm256_setzero_si256();
-  sums[6] = _mm256_setzero_si256();
-  sums[7] = _mm256_setzero_si256();
-
-  for (i = 0; i < 32; i++) {
-    __m256i r[8];
-
-    // load src and all ref[]
-    const __m256i s = _mm256_load_si256((const __m256i *)src_ptr);
-    r[0] = _mm256_loadu_si256((const __m256i *)&ref_ptr[0]);
-    r[1] = _mm256_loadu_si256((const __m256i *)&ref_ptr[1]);
-    r[2] = _mm256_loadu_si256((const __m256i *)&ref_ptr[2]);
-    r[3] = _mm256_loadu_si256((const __m256i *)&ref_ptr[3]);
-    r[4] = _mm256_loadu_si256((const __m256i *)&ref_ptr[4]);
-    r[5] = _mm256_loadu_si256((const __m256i *)&ref_ptr[5]);
-    r[6] = _mm256_loadu_si256((const __m256i *)&ref_ptr[6]);
-    r[7] = _mm256_loadu_si256((const __m256i *)&ref_ptr[7]);
-
-    // sum of the absolute differences between every ref[] to src
-    r[0] = _mm256_sad_epu8(r[0], s);
-    r[1] = _mm256_sad_epu8(r[1], s);
-    r[2] = _mm256_sad_epu8(r[2], s);
-    r[3] = _mm256_sad_epu8(r[3], s);
-    r[4] = _mm256_sad_epu8(r[4], s);
-    r[5] = _mm256_sad_epu8(r[5], s);
-    r[6] = _mm256_sad_epu8(r[6], s);
-    r[7] = _mm256_sad_epu8(r[7], s);
-
-    // sum every ref[]
-    sums[0] = _mm256_add_epi32(sums[0], r[0]);
-    sums[1] = _mm256_add_epi32(sums[1], r[1]);
-    sums[2] = _mm256_add_epi32(sums[2], r[2]);
-    sums[3] = _mm256_add_epi32(sums[3], r[3]);
-    sums[4] = _mm256_add_epi32(sums[4], r[4]);
-    sums[5] = _mm256_add_epi32(sums[5], r[5]);
-    sums[6] = _mm256_add_epi32(sums[6], r[6]);
-    sums[7] = _mm256_add_epi32(sums[7], r[7]);
-
-    src_ptr += src_stride;
-    ref_ptr += ref_stride;
-  }
-
-  calc_final_4(sums, sad_array);
-  calc_final_4(sums + 4, sad_array + 4);
-}
-
 void vpx_sad64x64x4d_avx2(const uint8_t *src_ptr, int src_stride,
                           const uint8_t *const ref_array[4], int ref_stride,
                           uint32_t sad_array[4]) {
diff --git a/vpx_dsp/x86/sad_sse3.asm b/vpx_dsp/x86/sad_sse3.asm
deleted file mode 100644
index acbd2e4fa..000000000
--- a/vpx_dsp/x86/sad_sse3.asm
+++ /dev/null
@@ -1,376 +0,0 @@
-;
-;  Copyright (c) 2010 The WebM project authors. All Rights Reserved.
-;
-;  Use of this source code is governed by a BSD-style license
-;  that can be found in the LICENSE file in the root of the source
-;  tree. An additional intellectual property rights grant can be found
-;  in the file PATENTS.  All contributing project authors may
-;  be found in the AUTHORS file in the root of the source tree.
-;
-
-%include "vpx_ports/x86_abi_support.asm"
-
-%macro STACK_FRAME_CREATE_X3 0
-%if ABI_IS_32BIT
-  %define     src_ptr       rsi
-  %define     src_stride    rax
-  %define     ref_ptr       rdi
-  %define     ref_stride    rdx
-  %define     end_ptr       rcx
-  %define     ret_var       rbx
-  %define     result_ptr    arg(4)
-  %define     height        dword ptr arg(4)
-    push        rbp
-    mov         rbp,        rsp
-    push        rsi
-    push        rdi
-    push        rbx
-
-    mov         rsi,        arg(0)              ; src_ptr
-    mov         rdi,        arg(2)              ; ref_ptr
-
-    movsxd      rax,        dword ptr arg(1)    ; src_stride
-    movsxd      rdx,        dword ptr arg(3)    ; ref_stride
-%else
-  %if LIBVPX_YASM_WIN64
-    SAVE_XMM 7, u
-    %define     src_ptr     rcx
-    %define     src_stride  rdx
-    %define     ref_ptr     r8
-    %define     ref_stride  r9
-    %define     end_ptr     r10
-    %define     ret_var     r11
-    %define     result_ptr  [rsp+xmm_stack_space+8+4*8]
-    %define     height      dword ptr [rsp+xmm_stack_space+8+4*8]
-  %else
-    %define     src_ptr     rdi
-    %define     src_stride  rsi
-    %define     ref_ptr     rdx
-    %define     ref_stride  rcx
-    %define     end_ptr     r9
-    %define     ret_var     r10
-    %define     result_ptr  r8
-    %define     height      r8
-  %endif
-%endif
-
-%endmacro
-
-%macro STACK_FRAME_DESTROY_X3 0
-  %define     src_ptr
-  %define     src_stride
-  %define     ref_ptr
-  %define     ref_stride
-  %define     end_ptr
-  %define     ret_var
-  %define     result_ptr
-  %define     height
-
-%if ABI_IS_32BIT
-    pop         rbx
-    pop         rdi
-    pop         rsi
-    pop         rbp
-%else
-  %if LIBVPX_YASM_WIN64
-    RESTORE_XMM
-  %endif
-%endif
-    ret
-%endmacro
-
-%macro PROCESS_16X2X3 5
-%if %1==0
-        movdqa          xmm0,       XMMWORD PTR [%2]
-        lddqu           xmm5,       XMMWORD PTR [%3]
-        lddqu           xmm6,       XMMWORD PTR [%3+1]
-        lddqu           xmm7,       XMMWORD PTR [%3+2]
-
-        psadbw          xmm5,       xmm0
-        psadbw          xmm6,       xmm0
-        psadbw          xmm7,       xmm0
-%else
-        movdqa          xmm0,       XMMWORD PTR [%2]
-        lddqu           xmm1,       XMMWORD PTR [%3]
-        lddqu           xmm2,       XMMWORD PTR [%3+1]
-        lddqu           xmm3,       XMMWORD PTR [%3+2]
-
-        psadbw          xmm1,       xmm0
-        psadbw          xmm2,       xmm0
-        psadbw          xmm3,       xmm0
-
-        paddw           xmm5,       xmm1
-        paddw           xmm6,       xmm2
-        paddw           xmm7,       xmm3
-%endif
-        movdqa          xmm0,       XMMWORD PTR [%2+%4]
-        lddqu           xmm1,       XMMWORD PTR [%3+%5]
-        lddqu           xmm2,       XMMWORD PTR [%3+%5+1]
-        lddqu           xmm3,       XMMWORD PTR [%3+%5+2]
-
-%if %1==0 || %1==1
-        lea             %2,         [%2+%4*2]
-        lea             %3,         [%3+%5*2]
-%endif
-
-        psadbw          xmm1,       xmm0
-        psadbw          xmm2,       xmm0
-        psadbw          xmm3,       xmm0
-
-        paddw           xmm5,       xmm1
-        paddw           xmm6,       xmm2
-        paddw           xmm7,       xmm3
-%endmacro
-
-%macro PROCESS_8X2X3 5
-%if %1==0
-        movq            mm0,       QWORD PTR [%2]
-        movq            mm5,       QWORD PTR [%3]
-        movq            mm6,       QWORD PTR [%3+1]
-        movq            mm7,       QWORD PTR [%3+2]
-
-        psadbw          mm5,       mm0
-        psadbw          mm6,       mm0
-        psadbw          mm7,       mm0
-%else
-        movq            mm0,       QWORD PTR [%2]
-        movq            mm1,       QWORD PTR [%3]
-        movq            mm2,       QWORD PTR [%3+1]
-        movq            mm3,       QWORD PTR [%3+2]
-
-        psadbw          mm1,       mm0
-        psadbw          mm2,       mm0
-        psadbw          mm3,       mm0
-
-        paddw           mm5,       mm1
-        paddw           mm6,       mm2
-        paddw           mm7,       mm3
-%endif
-        movq            mm0,       QWORD PTR [%2+%4]
-        movq            mm1,       QWORD PTR [%3+%5]
-        movq            mm2,       QWORD PTR [%3+%5+1]
-        movq            mm3,       QWORD PTR [%3+%5+2]
-
-%if %1==0 || %1==1
-        lea             %2,        [%2+%4*2]
-        lea             %3,        [%3+%5*2]
-%endif
-
-        psadbw          mm1,       mm0
-        psadbw          mm2,       mm0
-        psadbw          mm3,       mm0
-
-        paddw           mm5,       mm1
-        paddw           mm6,       mm2
-        paddw           mm7,       mm3
-%endmacro
-
-SECTION .text
-
-;void int vpx_sad16x16x3_sse3(
-;    unsigned char *src_ptr,
-;    int  src_stride,
-;    unsigned char *ref_ptr,
-;    int  ref_stride,
-;    int  *results)
-globalsym(vpx_sad16x16x3_sse3)
-sym(vpx_sad16x16x3_sse3):
-
-    STACK_FRAME_CREATE_X3
-
-        PROCESS_16X2X3 0, src_ptr, ref_ptr, src_stride, ref_stride
-        PROCESS_16X2X3 1, src_ptr, ref_ptr, src_stride, ref_stride
-        PROCESS_16X2X3 1, src_ptr, ref_ptr, src_stride, ref_stride
-        PROCESS_16X2X3 1, src_ptr, ref_ptr, src_stride, ref_stride
-        PROCESS_16X2X3 1, src_ptr, ref_ptr, src_stride, ref_stride
-        PROCESS_16X2X3 1, src_ptr, ref_ptr, src_stride, ref_stride
-        PROCESS_16X2X3 1, src_ptr, ref_ptr, src_stride, ref_stride
-        PROCESS_16X2X3 2, src_ptr, ref_ptr, src_stride, ref_stride
-
-        mov             rcx,        result_ptr
-
-        movq            xmm0,       xmm5
-        psrldq          xmm5,       8
-
-        paddw           xmm0,       xmm5
-        movd            [rcx],      xmm0
-;-
-        movq            xmm0,       xmm6
-        psrldq          xmm6,       8
-
-        paddw           xmm0,       xmm6
-        movd            [rcx+4],    xmm0
-;-
-        movq            xmm0,       xmm7
-        psrldq          xmm7,       8
-
-        paddw           xmm0,       xmm7
-        movd            [rcx+8],    xmm0
-
-    STACK_FRAME_DESTROY_X3
-
-;void int vpx_sad16x8x3_sse3(
-;    unsigned char *src_ptr,
-;    int  src_stride,
-;    unsigned char *ref_ptr,
-;    int  ref_stride,
-;    int  *results)
-globalsym(vpx_sad16x8x3_sse3)
-sym(vpx_sad16x8x3_sse3):
-
-    STACK_FRAME_CREATE_X3
-
-        PROCESS_16X2X3 0, src_ptr, ref_ptr, src_stride, ref_stride
-        PROCESS_16X2X3 1, src_ptr, ref_ptr, src_stride, ref_stride
-        PROCESS_16X2X3 1, src_ptr, ref_ptr, src_stride, ref_stride
-        PROCESS_16X2X3 2, src_ptr, ref_ptr, src_stride, ref_stride
-
-        mov             rcx,        result_ptr
-
-        movq            xmm0,       xmm5
-        psrldq          xmm5,       8
-
-        paddw           xmm0,       xmm5
-        movd            [rcx],      xmm0
-;-
-        movq            xmm0,       xmm6
-        psrldq          xmm6,       8
-
-        paddw           xmm0,       xmm6
-        movd            [rcx+4],    xmm0
-;-
-        movq            xmm0,       xmm7
-        psrldq          xmm7,       8
-
-        paddw           xmm0,       xmm7
-        movd            [rcx+8],    xmm0
-
-    STACK_FRAME_DESTROY_X3
-
-;void int vpx_sad8x16x3_sse3(
-;    unsigned char *src_ptr,
-;    int  src_stride,
-;    unsigned char *ref_ptr,
-;    int  ref_stride,
-;    int  *results)
-globalsym(vpx_sad8x16x3_sse3)
-sym(vpx_sad8x16x3_sse3):
-
-    STACK_FRAME_CREATE_X3
-
-        PROCESS_8X2X3 0, src_ptr, ref_ptr, src_stride, ref_stride
-        PROCESS_8X2X3 1, src_ptr, ref_ptr, src_stride, ref_stride
-        PROCESS_8X2X3 1, src_ptr, ref_ptr, src_stride, ref_stride
-        PROCESS_8X2X3 1, src_ptr, ref_ptr, src_stride, ref_stride
-        PROCESS_8X2X3 1, src_ptr, ref_ptr, src_stride, ref_stride
-        PROCESS_8X2X3 1, src_ptr, ref_ptr, src_stride, ref_stride
-        PROCESS_8X2X3 1, src_ptr, ref_ptr, src_stride, ref_stride
-        PROCESS_8X2X3 2, src_ptr, ref_ptr, src_stride, ref_stride
-
-        mov             rcx,        result_ptr
-
-        punpckldq       mm5,        mm6
-
-        movq            [rcx],      mm5
-        movd            [rcx+8],    mm7
-
-    STACK_FRAME_DESTROY_X3
-
-;void int vpx_sad8x8x3_sse3(
-;    unsigned char *src_ptr,
-;    int  src_stride,
-;    unsigned char *ref_ptr,
-;    int  ref_stride,
-;    int  *results)
-globalsym(vpx_sad8x8x3_sse3)
-sym(vpx_sad8x8x3_sse3):
-
-    STACK_FRAME_CREATE_X3
-
-        PROCESS_8X2X3 0, src_ptr, ref_ptr, src_stride, ref_stride
-        PROCESS_8X2X3 1, src_ptr, ref_ptr, src_stride, ref_stride
-        PROCESS_8X2X3 1, src_ptr, ref_ptr, src_stride, ref_stride
-        PROCESS_8X2X3 2, src_ptr, ref_ptr, src_stride, ref_stride
-
-        mov             rcx,        result_ptr
-
-        punpckldq       mm5,        mm6
-
-        movq            [rcx],      mm5
-        movd            [rcx+8],    mm7
-
-    STACK_FRAME_DESTROY_X3
-
-;void int vpx_sad4x4x3_sse3(
-;    unsigned char *src_ptr,
-;    int  src_stride,
-;    unsigned char *ref_ptr,
-;    int  ref_stride,
-;    int  *results)
-globalsym(vpx_sad4x4x3_sse3)
-sym(vpx_sad4x4x3_sse3):
-
-    STACK_FRAME_CREATE_X3
-
-        movd            mm0,        DWORD PTR [src_ptr]
-        movd            mm1,        DWORD PTR [ref_ptr]
-
-        movd            mm2,        DWORD PTR [src_ptr+src_stride]
-        movd            mm3,        DWORD PTR [ref_ptr+ref_stride]
-
-        punpcklbw       mm0,        mm2
-        punpcklbw       mm1,        mm3
-
-        movd            mm4,        DWORD PTR [ref_ptr+1]
-        movd            mm5,        DWORD PTR [ref_ptr+2]
-
-        movd            mm2,        DWORD PTR [ref_ptr+ref_stride+1]
-        movd            mm3,        DWORD PTR [ref_ptr+ref_stride+2]
-
-        psadbw          mm1,        mm0
-
-        punpcklbw       mm4,        mm2
-        punpcklbw       mm5,        mm3
-
-        psadbw          mm4,        mm0
-        psadbw          mm5,        mm0
-
-        lea             src_ptr,    [src_ptr+src_stride*2]
-        lea             ref_ptr,    [ref_ptr+ref_stride*2]
-
-        movd            mm0,        DWORD PTR [src_ptr]
-        movd            mm2,        DWORD PTR [ref_ptr]
-
-        movd            mm3,        DWORD PTR [src_ptr+src_stride]
-        movd            mm6,        DWORD PTR [ref_ptr+ref_stride]
-
-        punpcklbw       mm0,        mm3
-        punpcklbw       mm2,        mm6
-
-        movd            mm3,        DWORD PTR [ref_ptr+1]
-        movd            mm7,        DWORD PTR [ref_ptr+2]
-
-        psadbw          mm2,        mm0
-
-        paddw           mm1,        mm2
-
-        movd            mm2,        DWORD PTR [ref_ptr+ref_stride+1]
-        movd            mm6,        DWORD PTR [ref_ptr+ref_stride+2]
-
-        punpcklbw       mm3,        mm2
-        punpcklbw       mm7,        mm6
-
-        psadbw          mm3,        mm0
-        psadbw          mm7,        mm0
-
-        paddw           mm3,        mm4
-        paddw           mm7,        mm5
-
-        mov             rcx,        result_ptr
-
-        punpckldq       mm1,        mm3
-
-        movq            [rcx],      mm1
-        movd            [rcx+8],    mm7
-
-    STACK_FRAME_DESTROY_X3
diff --git a/vpx_dsp/x86/sad_sse4.asm b/vpx_dsp/x86/sad_sse4.asm
deleted file mode 100644
index 0818ed5f0..000000000
--- a/vpx_dsp/x86/sad_sse4.asm
+++ /dev/null
@@ -1,361 +0,0 @@
-;
-;  Copyright (c) 2010 The WebM project authors. All Rights Reserved.
-;
-;  Use of this source code is governed by a BSD-style license
-;  that can be found in the LICENSE file in the root of the source
-;  tree. An additional intellectual property rights grant can be found
-;  in the file PATENTS.  All contributing project authors may
-;  be found in the AUTHORS file in the root of the source tree.
-;
-
-
-%include "vpx_ports/x86_abi_support.asm"
-
-%macro PROCESS_16X2X8 1
-%if %1
-        movdqa          xmm0,       XMMWORD PTR [rsi]
-        movq            xmm1,       MMWORD PTR [rdi]
-        movq            xmm3,       MMWORD PTR [rdi+8]
-        movq            xmm2,       MMWORD PTR [rdi+16]
-        punpcklqdq      xmm1,       xmm3
-        punpcklqdq      xmm3,       xmm2
-
-        movdqa          xmm2,       xmm1
-        mpsadbw         xmm1,       xmm0,  0x0
-        mpsadbw         xmm2,       xmm0,  0x5
-
-        psrldq          xmm0,       8
-
-        movdqa          xmm4,       xmm3
-        mpsadbw         xmm3,       xmm0,  0x0
-        mpsadbw         xmm4,       xmm0,  0x5
-
-        paddw           xmm1,       xmm2
-        paddw           xmm1,       xmm3
-        paddw           xmm1,       xmm4
-%else
-        movdqa          xmm0,       XMMWORD PTR [rsi]
-        movq            xmm5,       MMWORD PTR [rdi]
-        movq            xmm3,       MMWORD PTR [rdi+8]
-        movq            xmm2,       MMWORD PTR [rdi+16]
-        punpcklqdq      xmm5,       xmm3
-        punpcklqdq      xmm3,       xmm2
-
-        movdqa          xmm2,       xmm5
-        mpsadbw         xmm5,       xmm0,  0x0
-        mpsadbw         xmm2,       xmm0,  0x5
-
-        psrldq          xmm0,       8
-
-        movdqa          xmm4,       xmm3
-        mpsadbw         xmm3,       xmm0,  0x0
-        mpsadbw         xmm4,       xmm0,  0x5
-
-        paddw           xmm5,       xmm2
-        paddw           xmm5,       xmm3
-        paddw           xmm5,       xmm4
-
-        paddw           xmm1,       xmm5
-%endif
-        movdqa          xmm0,       XMMWORD PTR [rsi + rax]
-        movq            xmm5,       MMWORD PTR [rdi+ rdx]
-        movq            xmm3,       MMWORD PTR [rdi+ rdx+8]
-        movq            xmm2,       MMWORD PTR [rdi+ rdx+16]
-        punpcklqdq      xmm5,       xmm3
-        punpcklqdq      xmm3,       xmm2
-
-        lea             rsi,        [rsi+rax*2]
-        lea             rdi,        [rdi+rdx*2]
-
-        movdqa          xmm2,       xmm5
-        mpsadbw         xmm5,       xmm0,  0x0
-        mpsadbw         xmm2,       xmm0,  0x5
-
-        psrldq          xmm0,       8
-        movdqa          xmm4,       xmm3
-        mpsadbw         xmm3,       xmm0,  0x0
-        mpsadbw         xmm4,       xmm0,  0x5
-
-        paddw           xmm5,       xmm2
-        paddw           xmm5,       xmm3
-        paddw           xmm5,       xmm4
-
-        paddw           xmm1,       xmm5
-%endmacro
-
-%macro PROCESS_8X2X8 1
-%if %1
-        movq            xmm0,       MMWORD PTR [rsi]
-        movq            xmm1,       MMWORD PTR [rdi]
-        movq            xmm3,       MMWORD PTR [rdi+8]
-        punpcklqdq      xmm1,       xmm3
-
-        movdqa          xmm2,       xmm1
-        mpsadbw         xmm1,       xmm0,  0x0
-        mpsadbw         xmm2,       xmm0,  0x5
-        paddw           xmm1,       xmm2
-%else
-        movq            xmm0,       MMWORD PTR [rsi]
-        movq            xmm5,       MMWORD PTR [rdi]
-        movq            xmm3,       MMWORD PTR [rdi+8]
-        punpcklqdq      xmm5,       xmm3
-
-        movdqa          xmm2,       xmm5
-        mpsadbw         xmm5,       xmm0,  0x0
-        mpsadbw         xmm2,       xmm0,  0x5
-        paddw           xmm5,       xmm2
-
-        paddw           xmm1,       xmm5
-%endif
-        movq            xmm0,       MMWORD PTR [rsi + rax]
-        movq            xmm5,       MMWORD PTR [rdi+ rdx]
-        movq            xmm3,       MMWORD PTR [rdi+ rdx+8]
-        punpcklqdq      xmm5,       xmm3
-
-        lea             rsi,        [rsi+rax*2]
-        lea             rdi,        [rdi+rdx*2]
-
-        movdqa          xmm2,       xmm5
-        mpsadbw         xmm5,       xmm0,  0x0
-        mpsadbw         xmm2,       xmm0,  0x5
-        paddw           xmm5,       xmm2
-
-        paddw           xmm1,       xmm5
-%endmacro
-
-%macro PROCESS_4X2X8 1
-%if %1
-        movd            xmm0,       [rsi]
-        movq            xmm1,       MMWORD PTR [rdi]
-        movq            xmm3,       MMWORD PTR [rdi+8]
-        punpcklqdq      xmm1,       xmm3
-
-        mpsadbw         xmm1,       xmm0,  0x0
-%else
-        movd            xmm0,       [rsi]
-        movq            xmm5,       MMWORD PTR [rdi]
-        movq            xmm3,       MMWORD PTR [rdi+8]
-        punpcklqdq      xmm5,       xmm3
-
-        mpsadbw         xmm5,       xmm0,  0x0
-
-        paddw           xmm1,       xmm5
-%endif
-        movd            xmm0,       [rsi + rax]
-        movq            xmm5,       MMWORD PTR [rdi+ rdx]
-        movq            xmm3,       MMWORD PTR [rdi+ rdx+8]
-        punpcklqdq      xmm5,       xmm3
-
-        lea             rsi,        [rsi+rax*2]
-        lea             rdi,        [rdi+rdx*2]
-
-        mpsadbw         xmm5,       xmm0,  0x0
-
-        paddw           xmm1,       xmm5
-%endmacro
-
-%macro WRITE_AS_INTS 0
-    mov             rdi,        arg(4)           ;Results
-    pxor            xmm0, xmm0
-    movdqa          xmm2, xmm1
-    punpcklwd       xmm1, xmm0
-    punpckhwd       xmm2, xmm0
-
-    movdqa          [rdi],    xmm1
-    movdqa          [rdi + 16],    xmm2
-%endmacro
-
-SECTION .text
-
-;void vpx_sad16x16x8_sse4_1(
-;    const unsigned char *src_ptr,
-;    int  src_stride,
-;    const unsigned char *ref_ptr,
-;    int  ref_stride,
-;    unsigned short *sad_array);
-globalsym(vpx_sad16x16x8_sse4_1)
-sym(vpx_sad16x16x8_sse4_1):
-    push        rbp
-    mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 5
-    push        rsi
-    push        rdi
-    ; end prolog
-
-    mov             rsi,        arg(0)           ;src_ptr
-    mov             rdi,        arg(2)           ;ref_ptr
-
-    movsxd          rax,        dword ptr arg(1) ;src_stride
-    movsxd          rdx,        dword ptr arg(3) ;ref_stride
-
-    PROCESS_16X2X8 1
-    PROCESS_16X2X8 0
-    PROCESS_16X2X8 0
-    PROCESS_16X2X8 0
-    PROCESS_16X2X8 0
-    PROCESS_16X2X8 0
-    PROCESS_16X2X8 0
-    PROCESS_16X2X8 0
-
-    WRITE_AS_INTS
-
-    ; begin epilog
-    pop         rdi
-    pop         rsi
-    UNSHADOW_ARGS
-    pop         rbp
-    ret
-
-
-;void vpx_sad16x8x8_sse4_1(
-;    const unsigned char *src_ptr,
-;    int  src_stride,
-;    const unsigned char *ref_ptr,
-;    int  ref_stride,
-;    unsigned short *sad_array
-;);
-globalsym(vpx_sad16x8x8_sse4_1)
-sym(vpx_sad16x8x8_sse4_1):
-    push        rbp
-    mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 5
-    push        rsi
-    push        rdi
-    ; end prolog
-
-    mov             rsi,        arg(0)           ;src_ptr
-    mov             rdi,        arg(2)           ;ref_ptr
-
-    movsxd          rax,        dword ptr arg(1) ;src_stride
-    movsxd          rdx,        dword ptr arg(3) ;ref_stride
-
-    PROCESS_16X2X8 1
-    PROCESS_16X2X8 0
-    PROCESS_16X2X8 0
-    PROCESS_16X2X8 0
-
-    WRITE_AS_INTS
-
-    ; begin epilog
-    pop         rdi
-    pop         rsi
-    UNSHADOW_ARGS
-    pop         rbp
-    ret
-
-
-;void vpx_sad8x8x8_sse4_1(
-;    const unsigned char *src_ptr,
-;    int  src_stride,
-;    const unsigned char *ref_ptr,
-;    int  ref_stride,
-;    unsigned short *sad_array
-;);
-globalsym(vpx_sad8x8x8_sse4_1)
-sym(vpx_sad8x8x8_sse4_1):
-    push        rbp
-    mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 5
-    push        rsi
-    push        rdi
-    ; end prolog
-
-    mov             rsi,        arg(0)           ;src_ptr
-    mov             rdi,        arg(2)           ;ref_ptr
-
-    movsxd          rax,        dword ptr arg(1) ;src_stride
-    movsxd          rdx,        dword ptr arg(3) ;ref_stride
-
-    PROCESS_8X2X8 1
-    PROCESS_8X2X8 0
-    PROCESS_8X2X8 0
-    PROCESS_8X2X8 0
-
-    WRITE_AS_INTS
-
-    ; begin epilog
-    pop         rdi
-    pop         rsi
-    UNSHADOW_ARGS
-    pop         rbp
-    ret
-
-
-;void vpx_sad8x16x8_sse4_1(
-;    const unsigned char *src_ptr,
-;    int  src_stride,
-;    const unsigned char *ref_ptr,
-;    int  ref_stride,
-;    unsigned short *sad_array
-;);
-globalsym(vpx_sad8x16x8_sse4_1)
-sym(vpx_sad8x16x8_sse4_1):
-    push        rbp
-    mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 5
-    push        rsi
-    push        rdi
-    ; end prolog
-
-    mov             rsi,        arg(0)           ;src_ptr
-    mov             rdi,        arg(2)           ;ref_ptr
-
-    movsxd          rax,        dword ptr arg(1) ;src_stride
-    movsxd          rdx,        dword ptr arg(3) ;ref_stride
-
-    PROCESS_8X2X8 1
-    PROCESS_8X2X8 0
-    PROCESS_8X2X8 0
-    PROCESS_8X2X8 0
-    PROCESS_8X2X8 0
-    PROCESS_8X2X8 0
-    PROCESS_8X2X8 0
-    PROCESS_8X2X8 0
-
-    WRITE_AS_INTS
-
-    ; begin epilog
-    pop         rdi
-    pop         rsi
-    UNSHADOW_ARGS
-    pop         rbp
-    ret
-
-
-;void vpx_sad4x4x8_sse4_1(
-;    const unsigned char *src_ptr,
-;    int  src_stride,
-;    const unsigned char *ref_ptr,
-;    int  ref_stride,
-;    unsigned short *sad_array
-;);
-globalsym(vpx_sad4x4x8_sse4_1)
-sym(vpx_sad4x4x8_sse4_1):
-    push        rbp
-    mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 5
-    push        rsi
-    push        rdi
-    ; end prolog
-
-    mov             rsi,        arg(0)           ;src_ptr
-    mov             rdi,        arg(2)           ;ref_ptr
-
-    movsxd          rax,        dword ptr arg(1) ;src_stride
-    movsxd          rdx,        dword ptr arg(3) ;ref_stride
-
-    PROCESS_4X2X8 1
-    PROCESS_4X2X8 0
-
-    WRITE_AS_INTS
-
-    ; begin epilog
-    pop         rdi
-    pop         rsi
-    UNSHADOW_ARGS
-    pop         rbp
-    ret
-
-
-
-
diff --git a/vpx_dsp/x86/sad_ssse3.asm b/vpx_dsp/x86/sad_ssse3.asm
deleted file mode 100644
index a5bc6d730..000000000
--- a/vpx_dsp/x86/sad_ssse3.asm
+++ /dev/null
@@ -1,372 +0,0 @@
-;
-;  Copyright (c) 2010 The WebM project authors. All Rights Reserved.
-;
-;  Use of this source code is governed by a BSD-style license
-;  that can be found in the LICENSE file in the root of the source
-;  tree. An additional intellectual property rights grant can be found
-;  in the file PATENTS.  All contributing project authors may
-;  be found in the AUTHORS file in the root of the source tree.
-;
-
-
-%include "vpx_ports/x86_abi_support.asm"
-
-%macro PROCESS_16X2X3 1
-%if %1
-        movdqa          xmm0,       XMMWORD PTR [rsi]
-        lddqu           xmm5,       XMMWORD PTR [rdi]
-        lddqu           xmm6,       XMMWORD PTR [rdi+1]
-        lddqu           xmm7,       XMMWORD PTR [rdi+2]
-
-        psadbw          xmm5,       xmm0
-        psadbw          xmm6,       xmm0
-        psadbw          xmm7,       xmm0
-%else
-        movdqa          xmm0,       XMMWORD PTR [rsi]
-        lddqu           xmm1,       XMMWORD PTR [rdi]
-        lddqu           xmm2,       XMMWORD PTR [rdi+1]
-        lddqu           xmm3,       XMMWORD PTR [rdi+2]
-
-        psadbw          xmm1,       xmm0
-        psadbw          xmm2,       xmm0
-        psadbw          xmm3,       xmm0
-
-        paddw           xmm5,       xmm1
-        paddw           xmm6,       xmm2
-        paddw           xmm7,       xmm3
-%endif
-        movdqa          xmm0,       XMMWORD PTR [rsi+rax]
-        lddqu           xmm1,       XMMWORD PTR [rdi+rdx]
-        lddqu           xmm2,       XMMWORD PTR [rdi+rdx+1]
-        lddqu           xmm3,       XMMWORD PTR [rdi+rdx+2]
-
-        lea             rsi,        [rsi+rax*2]
-        lea             rdi,        [rdi+rdx*2]
-
-        psadbw          xmm1,       xmm0
-        psadbw          xmm2,       xmm0
-        psadbw          xmm3,       xmm0
-
-        paddw           xmm5,       xmm1
-        paddw           xmm6,       xmm2
-        paddw           xmm7,       xmm3
-%endmacro
-
-%macro PROCESS_16X2X3_OFFSET 2
-%if %1
-        movdqa          xmm0,       XMMWORD PTR [rsi]
-        movdqa          xmm4,       XMMWORD PTR [rdi]
-        movdqa          xmm7,       XMMWORD PTR [rdi+16]
-
-        movdqa          xmm5,       xmm7
-        palignr         xmm5,       xmm4,       %2
-
-        movdqa          xmm6,       xmm7
-        palignr         xmm6,       xmm4,       (%2+1)
-
-        palignr         xmm7,       xmm4,       (%2+2)
-
-        psadbw          xmm5,       xmm0
-        psadbw          xmm6,       xmm0
-        psadbw          xmm7,       xmm0
-%else
-        movdqa          xmm0,       XMMWORD PTR [rsi]
-        movdqa          xmm4,       XMMWORD PTR [rdi]
-        movdqa          xmm3,       XMMWORD PTR [rdi+16]
-
-        movdqa          xmm1,       xmm3
-        palignr         xmm1,       xmm4,       %2
-
-        movdqa          xmm2,       xmm3
-        palignr         xmm2,       xmm4,       (%2+1)
-
-        palignr         xmm3,       xmm4,       (%2+2)
-
-        psadbw          xmm1,       xmm0
-        psadbw          xmm2,       xmm0
-        psadbw          xmm3,       xmm0
-
-        paddw           xmm5,       xmm1
-        paddw           xmm6,       xmm2
-        paddw           xmm7,       xmm3
-%endif
-        movdqa          xmm0,       XMMWORD PTR [rsi+rax]
-        movdqa          xmm4,       XMMWORD PTR [rdi+rdx]
-        movdqa          xmm3,       XMMWORD PTR [rdi+rdx+16]
-
-        movdqa          xmm1,       xmm3
-        palignr         xmm1,       xmm4,       %2
-
-        movdqa          xmm2,       xmm3
-        palignr         xmm2,       xmm4,       (%2+1)
-
-        palignr         xmm3,       xmm4,       (%2+2)
-
-        lea             rsi,        [rsi+rax*2]
-        lea             rdi,        [rdi+rdx*2]
-
-        psadbw          xmm1,       xmm0
-        psadbw          xmm2,       xmm0
-        psadbw          xmm3,       xmm0
-
-        paddw           xmm5,       xmm1
-        paddw           xmm6,       xmm2
-        paddw           xmm7,       xmm3
-%endmacro
-
-%macro PROCESS_16X16X3_OFFSET 2
-%2_aligned_by_%1:
-
-        sub             rdi,        %1
-
-        PROCESS_16X2X3_OFFSET 1, %1
-        PROCESS_16X2X3_OFFSET 0, %1
-        PROCESS_16X2X3_OFFSET 0, %1
-        PROCESS_16X2X3_OFFSET 0, %1
-        PROCESS_16X2X3_OFFSET 0, %1
-        PROCESS_16X2X3_OFFSET 0, %1
-        PROCESS_16X2X3_OFFSET 0, %1
-        PROCESS_16X2X3_OFFSET 0, %1
-
-        jmp             %2_store_off
-
-%endmacro
-
-%macro PROCESS_16X8X3_OFFSET 2
-%2_aligned_by_%1:
-
-        sub             rdi,        %1
-
-        PROCESS_16X2X3_OFFSET 1, %1
-        PROCESS_16X2X3_OFFSET 0, %1
-        PROCESS_16X2X3_OFFSET 0, %1
-        PROCESS_16X2X3_OFFSET 0, %1
-
-        jmp             %2_store_off
-
-%endmacro
-
-SECTION .text
-
-;void int vpx_sad16x16x3_ssse3(
-;    unsigned char *src_ptr,
-;    int  src_stride,
-;    unsigned char *ref_ptr,
-;    int  ref_stride,
-;    int  *results)
-globalsym(vpx_sad16x16x3_ssse3)
-sym(vpx_sad16x16x3_ssse3):
-    push        rbp
-    mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 5
-    SAVE_XMM 7
-    push        rsi
-    push        rdi
-    push        rcx
-    ; end prolog
-
-        mov             rsi,        arg(0) ;src_ptr
-        mov             rdi,        arg(2) ;ref_ptr
-
-        mov             rdx,        0xf
-        and             rdx,        rdi
-
-        jmp .vpx_sad16x16x3_ssse3_skiptable
-.vpx_sad16x16x3_ssse3_jumptable:
-        dd .vpx_sad16x16x3_ssse3_aligned_by_0  - .vpx_sad16x16x3_ssse3_do_jump
-        dd .vpx_sad16x16x3_ssse3_aligned_by_1  - .vpx_sad16x16x3_ssse3_do_jump
-        dd .vpx_sad16x16x3_ssse3_aligned_by_2  - .vpx_sad16x16x3_ssse3_do_jump
-        dd .vpx_sad16x16x3_ssse3_aligned_by_3  - .vpx_sad16x16x3_ssse3_do_jump
-        dd .vpx_sad16x16x3_ssse3_aligned_by_4  - .vpx_sad16x16x3_ssse3_do_jump
-        dd .vpx_sad16x16x3_ssse3_aligned_by_5  - .vpx_sad16x16x3_ssse3_do_jump
-        dd .vpx_sad16x16x3_ssse3_aligned_by_6  - .vpx_sad16x16x3_ssse3_do_jump
-        dd .vpx_sad16x16x3_ssse3_aligned_by_7  - .vpx_sad16x16x3_ssse3_do_jump
-        dd .vpx_sad16x16x3_ssse3_aligned_by_8  - .vpx_sad16x16x3_ssse3_do_jump
-        dd .vpx_sad16x16x3_ssse3_aligned_by_9  - .vpx_sad16x16x3_ssse3_do_jump
-        dd .vpx_sad16x16x3_ssse3_aligned_by_10 - .vpx_sad16x16x3_ssse3_do_jump
-        dd .vpx_sad16x16x3_ssse3_aligned_by_11 - .vpx_sad16x16x3_ssse3_do_jump
-        dd .vpx_sad16x16x3_ssse3_aligned_by_12 - .vpx_sad16x16x3_ssse3_do_jump
-        dd .vpx_sad16x16x3_ssse3_aligned_by_13 - .vpx_sad16x16x3_ssse3_do_jump
-        dd .vpx_sad16x16x3_ssse3_aligned_by_14 - .vpx_sad16x16x3_ssse3_do_jump
-        dd .vpx_sad16x16x3_ssse3_aligned_by_15 - .vpx_sad16x16x3_ssse3_do_jump
-.vpx_sad16x16x3_ssse3_skiptable:
-
-        call .vpx_sad16x16x3_ssse3_do_jump
-.vpx_sad16x16x3_ssse3_do_jump:
-        pop             rcx                         ; get the address of do_jump
-        mov             rax,  .vpx_sad16x16x3_ssse3_jumptable - .vpx_sad16x16x3_ssse3_do_jump
-        add             rax,  rcx  ; get the absolute address of vpx_sad16x16x3_ssse3_jumptable
-
-        movsxd          rax,  dword [rax + 4*rdx]   ; get the 32 bit offset from the jumptable
-        add             rcx,        rax
-
-        movsxd          rax,        dword ptr arg(1) ;src_stride
-        movsxd          rdx,        dword ptr arg(3) ;ref_stride
-
-        jmp             rcx
-
-        PROCESS_16X16X3_OFFSET 0,  .vpx_sad16x16x3_ssse3
-        PROCESS_16X16X3_OFFSET 1,  .vpx_sad16x16x3_ssse3
-        PROCESS_16X16X3_OFFSET 2,  .vpx_sad16x16x3_ssse3
-        PROCESS_16X16X3_OFFSET 3,  .vpx_sad16x16x3_ssse3
-        PROCESS_16X16X3_OFFSET 4,  .vpx_sad16x16x3_ssse3
-        PROCESS_16X16X3_OFFSET 5,  .vpx_sad16x16x3_ssse3
-        PROCESS_16X16X3_OFFSET 6,  .vpx_sad16x16x3_ssse3
-        PROCESS_16X16X3_OFFSET 7,  .vpx_sad16x16x3_ssse3
-        PROCESS_16X16X3_OFFSET 8,  .vpx_sad16x16x3_ssse3
-        PROCESS_16X16X3_OFFSET 9,  .vpx_sad16x16x3_ssse3
-        PROCESS_16X16X3_OFFSET 10, .vpx_sad16x16x3_ssse3
-        PROCESS_16X16X3_OFFSET 11, .vpx_sad16x16x3_ssse3
-        PROCESS_16X16X3_OFFSET 12, .vpx_sad16x16x3_ssse3
-        PROCESS_16X16X3_OFFSET 13, .vpx_sad16x16x3_ssse3
-        PROCESS_16X16X3_OFFSET 14, .vpx_sad16x16x3_ssse3
-
-.vpx_sad16x16x3_ssse3_aligned_by_15:
-        PROCESS_16X2X3 1
-        PROCESS_16X2X3 0
-        PROCESS_16X2X3 0
-        PROCESS_16X2X3 0
-        PROCESS_16X2X3 0
-        PROCESS_16X2X3 0
-        PROCESS_16X2X3 0
-        PROCESS_16X2X3 0
-
-.vpx_sad16x16x3_ssse3_store_off:
-        mov             rdi,        arg(4) ;Results
-
-        movq            xmm0,       xmm5
-        psrldq          xmm5,       8
-
-        paddw           xmm0,       xmm5
-        movd            [rdi],      xmm0
-;-
-        movq            xmm0,       xmm6
-        psrldq          xmm6,       8
-
-        paddw           xmm0,       xmm6
-        movd            [rdi+4],    xmm0
-;-
-        movq            xmm0,       xmm7
-        psrldq          xmm7,       8
-
-        paddw           xmm0,       xmm7
-        movd            [rdi+8],    xmm0
-
-    ; begin epilog
-    pop         rcx
-    pop         rdi
-    pop         rsi
-    RESTORE_XMM
-    UNSHADOW_ARGS
-    pop         rbp
-    ret
-
-;void int vpx_sad16x8x3_ssse3(
-;    unsigned char *src_ptr,
-;    int  src_stride,
-;    unsigned char *ref_ptr,
-;    int  ref_stride,
-;    int  *results)
-globalsym(vpx_sad16x8x3_ssse3)
-sym(vpx_sad16x8x3_ssse3):
-    push        rbp
-    mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 5
-    SAVE_XMM 7
-    push        rsi
-    push        rdi
-    push        rcx
-    ; end prolog
-
-        mov             rsi,        arg(0) ;src_ptr
-        mov             rdi,        arg(2) ;ref_ptr
-
-        mov             rdx,        0xf
-        and             rdx,        rdi
-
-        jmp .vpx_sad16x8x3_ssse3_skiptable
-.vpx_sad16x8x3_ssse3_jumptable:
-        dd .vpx_sad16x8x3_ssse3_aligned_by_0  - .vpx_sad16x8x3_ssse3_do_jump
-        dd .vpx_sad16x8x3_ssse3_aligned_by_1  - .vpx_sad16x8x3_ssse3_do_jump
-        dd .vpx_sad16x8x3_ssse3_aligned_by_2  - .vpx_sad16x8x3_ssse3_do_jump
-        dd .vpx_sad16x8x3_ssse3_aligned_by_3  - .vpx_sad16x8x3_ssse3_do_jump
-        dd .vpx_sad16x8x3_ssse3_aligned_by_4  - .vpx_sad16x8x3_ssse3_do_jump
-        dd .vpx_sad16x8x3_ssse3_aligned_by_5  - .vpx_sad16x8x3_ssse3_do_jump
-        dd .vpx_sad16x8x3_ssse3_aligned_by_6  - .vpx_sad16x8x3_ssse3_do_jump
-        dd .vpx_sad16x8x3_ssse3_aligned_by_7  - .vpx_sad16x8x3_ssse3_do_jump
-        dd .vpx_sad16x8x3_ssse3_aligned_by_8  - .vpx_sad16x8x3_ssse3_do_jump
-        dd .vpx_sad16x8x3_ssse3_aligned_by_9  - .vpx_sad16x8x3_ssse3_do_jump
-        dd .vpx_sad16x8x3_ssse3_aligned_by_10 - .vpx_sad16x8x3_ssse3_do_jump
-        dd .vpx_sad16x8x3_ssse3_aligned_by_11 - .vpx_sad16x8x3_ssse3_do_jump
-        dd .vpx_sad16x8x3_ssse3_aligned_by_12 - .vpx_sad16x8x3_ssse3_do_jump
-        dd .vpx_sad16x8x3_ssse3_aligned_by_13 - .vpx_sad16x8x3_ssse3_do_jump
-        dd .vpx_sad16x8x3_ssse3_aligned_by_14 - .vpx_sad16x8x3_ssse3_do_jump
-        dd .vpx_sad16x8x3_ssse3_aligned_by_15 - .vpx_sad16x8x3_ssse3_do_jump
-.vpx_sad16x8x3_ssse3_skiptable:
-
-        call .vpx_sad16x8x3_ssse3_do_jump
-.vpx_sad16x8x3_ssse3_do_jump:
-        pop             rcx                         ; get the address of do_jump
-        mov             rax,  .vpx_sad16x8x3_ssse3_jumptable - .vpx_sad16x8x3_ssse3_do_jump
-        add             rax,  rcx  ; get the absolute address of vpx_sad16x8x3_ssse3_jumptable
-
-        movsxd          rax,  dword [rax + 4*rdx]   ; get the 32 bit offset from the jumptable
-        add             rcx,        rax
-
-        movsxd          rax,        dword ptr arg(1) ;src_stride
-        movsxd          rdx,        dword ptr arg(3) ;ref_stride
-
-        jmp             rcx
-
-        PROCESS_16X8X3_OFFSET 0,  .vpx_sad16x8x3_ssse3
-        PROCESS_16X8X3_OFFSET 1,  .vpx_sad16x8x3_ssse3
-        PROCESS_16X8X3_OFFSET 2,  .vpx_sad16x8x3_ssse3
-        PROCESS_16X8X3_OFFSET 3,  .vpx_sad16x8x3_ssse3
-        PROCESS_16X8X3_OFFSET 4,  .vpx_sad16x8x3_ssse3
-        PROCESS_16X8X3_OFFSET 5,  .vpx_sad16x8x3_ssse3
-        PROCESS_16X8X3_OFFSET 6,  .vpx_sad16x8x3_ssse3
-        PROCESS_16X8X3_OFFSET 7,  .vpx_sad16x8x3_ssse3
-        PROCESS_16X8X3_OFFSET 8,  .vpx_sad16x8x3_ssse3
-        PROCESS_16X8X3_OFFSET 9,  .vpx_sad16x8x3_ssse3
-        PROCESS_16X8X3_OFFSET 10, .vpx_sad16x8x3_ssse3
-        PROCESS_16X8X3_OFFSET 11, .vpx_sad16x8x3_ssse3
-        PROCESS_16X8X3_OFFSET 12, .vpx_sad16x8x3_ssse3
-        PROCESS_16X8X3_OFFSET 13, .vpx_sad16x8x3_ssse3
-        PROCESS_16X8X3_OFFSET 14, .vpx_sad16x8x3_ssse3
-
-.vpx_sad16x8x3_ssse3_aligned_by_15:
-
-        PROCESS_16X2X3 1
-        PROCESS_16X2X3 0
-        PROCESS_16X2X3 0
-        PROCESS_16X2X3 0
-
-.vpx_sad16x8x3_ssse3_store_off:
-        mov             rdi,        arg(4) ;Results
-
-        movq            xmm0,       xmm5
-        psrldq          xmm5,       8
-
-        paddw           xmm0,       xmm5
-        movd            [rdi],      xmm0
-;-
-        movq            xmm0,       xmm6
-        psrldq          xmm6,       8
-
-        paddw           xmm0,       xmm6
-        movd            [rdi+4],    xmm0
-;-
-        movq            xmm0,       xmm7
-        psrldq          xmm7,       8
-
-        paddw           xmm0,       xmm7
-        movd            [rdi+8],    xmm0
-
-    ; begin epilog
-    pop         rcx
-    pop         rdi
-    pop         rsi
-    RESTORE_XMM
-    UNSHADOW_ARGS
-    pop         rbp
-    ret